Google AI Mode uses multi-step reasoning to synthesise responses from several sources simultaneously, while AI Overviews extract a single direct answer from one dominant result. For B2B content teams, this difference matters more than any single keyword tactic: structured data, entity-rich writing, and consistent brand citations across multiple authoritative pages now determine whether your brand earns a mention at all.

What Google AI Mode Actually is, and Why It Behaves Differently from AI Overviews

Most guides treat Google AI Mode as a more capable version of AI Overviews, which is not at all. The two features share the same interface but operate on fundamentally different architectures, and understanding the distinction is the first practical requirement for any content team building search visibility in 2026.

Google AI Overviews, which launched broadly in 2024, work through what practitioners have called single-source extraction. The system identifies the most authoritative ranked result for a query, pulls the most answer-ready passage from that page, and surfaces it at the top of the search results. The content strategy implication of this architecture was relatively familiar: write the clearest possible answer to the target question, structure it for easy extraction, and maintain high topical authority on the page. That framework produced measurable results, and it continues to do so for informational queries.

Google AI Mode operates on a different model entirely. Announced at Google I/O 2025 and expanded significantly through 2026, reaching over one billion users, AI Mode uses a technique the Google research team has described as query fan-out. When a user submits a query, AI Mode does not look for a single dominant answer. Instead, it decomposes the query into a set of related sub-questions, runs those sub-questions across the index simultaneously, and synthesises a response grounded in the most credible results it finds across multiple sources. The output is not an extracted answer. It is a generated synthesis, built from citations and shaping the questions for the users to select. It is similar to how NotebookLLM was developed to produce various content formats ranging from infographics, slide deck, mindmap, flash cards, reports, quiz, data tables, audio overview and video overview.

The shift from extraction to synthesis shows us that Google will keep switching their search interfaces, and no matter which brand or no brand will not get cited that easily. In an extraction model, the winner is the page with the best answer. In a synthesis model, the winners are the pages that collectively construct the most complete and credible response to a question and its related sub-questions. That distinction is the foundation of every content architecture decision that follows.

Try typing out your intended search keywords inside the AI Mode, and you will understand that its answers can appear inaccurate and untrue at times.

How the Mechanism Difference Reshapes Your Content Priorities

The practical consequence of the query fan-out approach is that a single well-optimised page is no longer sufficient to earn consistent AI Mode visibility. AI Overviews reward the page that answers the question most authoritatively. AI Mode rewards the brand that appears credibly across multiple pages that collectively address the question and its adjacent sub-questions.

This shift has two immediate implications for B2B content teams. The first is topical depth. In an AI Mode environment, a brand needs a cluster of credible, source-cited content across the full question space, not a single hero post optimised for one primary keyword. A B2B SaaS company trying to appear in AI Mode results for queries about their category will benefit more from five well-researched posts that collectively address different angles of the same problem than from one comprehensive guide designed to capture the featured snippet.

The second implication is entity recognition. AI Mode’s synthesis engine does not simply match keywords. It recognises entities: organisations, people, products, and frameworks that have been consistently represented across multiple credible sources. For B2B brands, this means the consistency of your brand name, your product names, and your key claims across your own content and in external mentions matters far more than keyword frequency on any single page. The brand that appears in third-party analyst commentary, case studies, and external press coverage is the brand that AI Mode learns to associate with the topic space. The brand that exists only on its own blog is largely invisible to the synthesis.

There is also a third implication that sits beneath both of these: sourcing discipline. AI Mode synthesises across sources, but it does not treat all sources equally. The synthesis engine applies credibility signals that weight verifiable claims, external source citations, and author authority. A brand that publishes content with traceable factual claims and authoritative external links earns higher synthesis weight than one that publishes general opinion without evidence. The practical consequence is that every post your team publishes is either building or undermining your AI Mode citation equity, depending on how well it documents its claims.

Why B2B Content Teams are Optimising for the Wrong Surface

Here is the tension most content teams are not naming clearly: the playbook that worked for AI Overviews, the featured-snippet-first, answer-ready, single-page architecture, is not wrong. It still applies. AI Overviews reaches an estimated 1+ billion users, and remains the dominant AI surface in Google Search. Abandoning that approach would be a mistake.

The mistake is treating it as the only strategy available.

B2B buyers who use AI Mode are typically in a different phase of their research as they are not looking for a quick definition. They are synthesising perspectives on a decision, comparing approaches, and building a mental model of a market before they engage with any one at all. AI Mode is where complex commercial intent lives. It is the surface most likely to influence shortlisting and purchase consideration for B2B products and services, precisely because the queries it handles are the multi-part exploratory ones that precede a decision.

If your 2026 content strategy is built entirely around capturing featured snippets and AI Overview answers, you are optimising for top-of-funnel informational queries and leaving the mid-funnel decision-making moments largely unaddressed. A brand that earns no citations in AI Mode outputs during the decision phase is a brand that does not exist for the buyer at the moment they are most ready to act.

The brands building topical authority clusters and consistent entity presence now are accumulating citation equity in an AI search environment that will only grow more competitive over the next 18 months.

A Practitioner’s Note: Why I Started Treating Content as an Infrastructure Platform, Not Just Publishing Packages

This is the part where I want to step out of the theory and speak from my own build experience.

When I started building LadyinTechverse in early 2025, I chose the harder route on purpose. I built it on self-hosted WordPress over my own hosting environment because I wanted control over the full foundation: the code, the database, the content, the brand architecture, and the direction of the platform.

It was never meant to be for blogging only, albeit the intention towards a full ownership decision.

If I was going to build a digital sanctuary around AI, digital transformation, marketing transformation, and strategic visibility, I did not want the whole foundation to depend on a platform I could not fully control. At the time, that felt like a practical technology choice. Now, in the AI search era, I see it differently. It was also an early decision about retrieval, entity control, and long-term brand memory.

The messy part came later, where my agents kept failing me because I did not provide sufficient constraints.

As LadyinTechverse grew, my content system started collecting layers of thinking across TextEdit notes, Obsidian dumps, Markdown files, JSON structures, drafts, SEO audits, image prompt libraries, brand rules, conceptual ideas, technical fixes, AI agent workflows and much much more. Individually, each piece made sense. Collectively, it became clear how easily a brand’s knowledge can become fragmented, even when the work itself is of high quality. Documentation can be a long term pain like having jumper’s knees.

That was the uncomfortable realisation: a content library is no longer just a place where published posts live. It is becoming the retrieval layer that AI systems, search engines, and answer engines use to understand whether your brand is coherent enough to cite.

It sounds technical, but the business implication is very simple. If your brand context is scattered, inconsistent, outdated, or poorly structured, AI systems will struggle to understand what your organisation actually stands for. They may miss you entirely. Worse, they may describe you inaccurately, confidently, and publicly like how I described them in my previous article, AI hallucination risk.

I saw this even more clearly when I tested the Claude Design System for branded content and production workflows. It was useful for broad outputs such as webpage concepts, visual directions, slide structures, and early creative artefact experiments. It gave me speed, but high quality was indeed compromised, and suffice to say that Claude should stick to coding themselves – allow third-party creative software integrations would create a much better deal for them than trying to monopolise the entire content publishing journey.

Where it became weaker was in preserving granular brand logic: colour hierarchy, reusable design rules, interface consistency, editorial judgement, naming conventions, verified claims, and the deeper strategic intent behind the content. The issue was not whether the tool was useful. It was whether the workflow could be gated, controlled, corrected, and trusted across repeated use.

That distinction matters for lean content teams. It is not uncommon to see workspaces fail because their tools are sitting on top of weak context layers without much guardrails, brand rules, structured content system and more. They have scattered documents, inconsistent messaging, old claims, duplicated pages, unclear ownership, and no single view of what the brand is allowed to say, prove, or be known for.

That is why I started building and refining my own AI-assisted workflows and LITV AI SEO Agent v2.0 logic around structure, auditability, and citation readiness. Not because I wanted to produce more content for the sake of content publishing. Frankly, I am already immersed enough in AI Fatigue Mode to know that more output is not the answer, hence I paused my social posts in the meantime to run some tests later on. Don’t be surprised if you see odd content across my socials ~ Lol!

I am not the only one here who was asked to rest by Claude and it also thinks that we are unusually eating one animal’s food.

What I have always needed was a system that could preserve context better across content, technical SEO, SXO, AEO, GEO, AI visibility checks, and brand consistency.

That is the shift I think B2B teams need to take seriously now. Google AI Mode visibility is not earned by publishing endlessly and hoping one page gets picked up. It is earned by building a content system that behaves like a manufacturing facility: consistent entities, clear internal links, source-backed claims, structured schema, named frameworks, updated pages, and a brand narrative that does not reset every time a new AI tool enters the workflow.

The real risk is assuming AI tools can literally understand your brand before you built the system that teaches them what your brand means and where its rooted.

Five Content Architecture Shifts for Google AI Mode Visibility

The answer is not to abandon your existing AI Overviews strategy. It is to extend it deliberately. The following five shifts are the practical additions that a B2B content team needs to make to earn consistent visibility across both surfaces.

The first shift is from keyword optimisation to entity consistency. Every piece of content your organisation publishes should refer to your brand, your flagship products, and your core frameworks by the same names, every time. AI Mode’s synthesis engine builds entity associations by recognising consistent co-occurrence patterns across credible sources. If you refer to your methodology as “the four-signal framework” in one post and “the attribution model” in another, the entity association never fully consolidates. Name your frameworks, use those names consistently, and link between every post that references them.

The second shift is from single-page depth to topical cluster density. Identify the primary question your brand should own in AI Mode results, then map the five to eight sub-questions that a query fan-out technique would decompose it into. If your brand should appear in AI Mode answers about B2B marketing strategy in the AI search era, the sub-questions include: what attribution metrics work in AI search, how does AI search affect content ROI, what frameworks exist for measuring brand visibility in AI results, and how should content teams structure for AI citations. You need credible, source-cited content that addresses each of these distinctly. The post on generative engine optimisation covers the citation signal principles that underpin this cluster architecture.

The third shift is from FAQ formatting to citation-worthy density. AI Mode does not simply extract FAQ answers. It looks for passages that are specific, verifiable, and supported by credible sources. The shift is from writing FAQ answers that pass the “clear and concise” test to writing FAQ answers that pass the “worth citing” test. A forty-word FAQ answer that references a named framework, includes a verifiable statistic, and links to an authoritative external source is substantially more likely to appear in an AI Mode synthesis than one that paraphrases the question in general terms. The principles behind this approach are covered in detail in the post on answer engine optimisation for B2B brands.

The fourth shift is from standard article schema to full entity markup. AI Mode benefits from structured data that tells Google not just what a page is about, but who produced it, what organisation they represent, what the article cites, and what claims it makes. BlogPosting schema is the floor. Author schema with sameAs properties linking to verified profiles, Organisation schema with consistent naming, and FAQPage schema where applicable are the incremental additions that improve entity legibility for the AI Mode synthesis engine and increase the probability of accurate citation.

The fifth shift is from publishing to verification. As AI search surfaces multiply, the risk of inaccurate citation compounds. Content that makes unverifiable claims or relies on statistics that cannot be traced to a public source is more likely to be filtered out of AI Mode synthesis, or worse, cited inaccurately in a way that damages rather than builds brand authority. Every factual claim in your content library should be traceable to a publicly accessible source. This is a visibility measure, and the case for content verification is detailed in the post on AI hallucination risk for B2B content teams.

Final Thoughts

Google AI Mode and AI Overviews are two distinct products operating on distinct architectures with distinct content requirements. The teams that understand this in 2026 will build content strategies that earn citations on both surfaces. The teams that treat them interchangeably will optimise for one, miss the other, and find themselves invisible in the AI search results that matter most to buyers at the decision stage.

The LITV AI SEO Agent 2.0 was built specifically to close this gap. It audits your existing content for entity consistency, topical cluster coverage, and AI Mode citation readiness, and surfaces the specific changes that will have the greatest impact on your visibility across both surfaces. The free audit is at seoagent.ladyintechverse.com.

Both AI surfaces reward the same underlying content discipline: verifiable claims, entity-consistent language, source-cited specificity, and a topical cluster deep enough to address not just the question your buyer is asking, but the five questions they will ask next. Build for both. Neither surface is optional now.

Google removed FAQ rich results from search on 7 May 2026. Blue links are being demoted and 60 percent of all Google searches now end without a single click. And in AI Mode, that figure rises to 93 percent. If your content is not being cited inside an AI-generated answer, it effectively does not exist for a growing proportion of your target audience. That shift makes AI hallucination the most consequential brand risk in B2B marketing right now, and because of these sudden changes, teams have no systematic process in place to manage it.

The zero-click reality and the hallucination problem are not separate issues. They are the same issue viewed from two different angles. When your brand’s best route to a senior buyer is through an AI-generated answer rather than a clicked blue link, the accuracy of what that AI says about your brand, your category, and your competitors becomes a direct commercial risk, not a technical curiosity.

The Search Landscape That Created This Risk

Google’s AI Mode and the End of the Informational Click

On 19 May 2026 at Google I/O, Google confirmed what its own data had been signalling for 18 months: AI Mode is now the primary search interface, not a parallel feature sitting alongside traditional results. The new search is built around AI Mode, conversational follow-ups, and autonomous information agents that monitor the web around the clock. For users, this means faster answers and fewer blue links.

The numbers behind that shift are stark. 60 percent of traditional Google searches end without a click. Eighty-three percent of searches that trigger AI Overviews end without a click. Ninety-three percent of searches in AI Mode end without a click. Only one percent of users click on links inside an AI Overview. For B2B marketing teams whose content existed primarily to move buyers down a consideration funnel through organic search traffic, this is a structural change to the fundamental distribution mechanism, not a temporary dip in click-through rates.

Google’s deprecation of FAQ rich results on 7 May 2026 sits inside the same strategic direction. Google is moving further towards AI-generated search experiences, including AI Overviews, while reducing the influence of some traditional rich result formats. That means brands need to think less about winning a bigger blue-link listing and more about how their content is understood, cited, and surfaced across modern search experiences.

FAQPage schema itself is not dead. The markup continues to be crawlable by AI retrieval systems, and it remains a meaningful signal for how AI search interfaces understand your content. The SERP dropdown feature is gone. The citation value remains. The distinction matters because it changes where your editorial investment should go: not into expanding your Google listing footprint, but into building content that earns citations in AI-generated answers.

The Second Dimension of Hallucination Risk: What AI Says About You

Most conversations about AI hallucination in marketing focus on what happens when the AI your team uses to write content makes up a statistic. That is a real and consequential risk, and the verification framework later in this post addresses it directly. But in a zero-click world, there is a second dimension of hallucination risk that B2B marketing leaders have been slower to address: what AI says about your brand when a buyer asks.

A recent study found that Google’s AI Overview offers correct and reputably sourced summaries nine out of ten times. But while 90 percent sounds like a passing grade, the failure rate adds up in a matter of minutes. In an internal analysis of Gemini 3, Google found that the AI model produced incorrect information 28 percent of the time. Google claims AI Overviews are more accurate because they draw on Google search results before answering.

When AI Overviews get things wrong, the incorrect answer can be traced to several issues. Sometimes the AI cited a website that could not back up the information. Other times, it cited a website with the correct information but got the information wrong. In some cases, the overview got the answer correct but then proceeded to provide additional context that was wrong.

For a B2B brand, each of those failure modes translates to a specific commercial consequence. A buyer asking an AI search interface about your product category, your pricing model, your leadership team, or your service scope may receive an answer that is plausible-sounding, confidently presented, and factually wrong about your organisation. They will not know it is wrong. They may never reach your website to discover the discrepancy. The AI answer is the first and only impression.

This is the visibility paradox that the LITV AI SEO Agent 2.0 was built to address directly. It audits your brand’s AI search visibility and citation accuracy across ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini, and Bing Copilot, surfacing exactly where AI systems are generating incorrect or absent information about your brand and what structured content fixes are required to close those gaps. Understanding what AI is currently saying about you is the necessary first step before any verification or content governance programme can be effective.

Let’s sidestep for a bit with Search browsers in 2026

One thing’s for sure – if you do not want to use Google Chrome browser anymore, you have the freedom to decide if you want to choose from these few browsers known for privacy and ad-free services:

• Brave (if you want top-tier privacy, aggressive ad blocking, and fast performance. Brave is developed by US-based Brave Software, Inc., owned by former Mozilla CEO) – hailed as one of the Top 3 browsers in 2026.
• Vivaldi (if you are a power user who wants extreme customisation, advanced tab management, and built-in productivity tools. It is based in Norway and is employee-owned) – hailed as one of the Top 3 browsers in 2026.
• Mozilla Firefox (existed way before Chrome – Firefox used to be my favourite browser back then if I had to choose between the good-riddance “Internet Explorer”, its nemesis )
• Zen Browser (a free and open-source project that is a modified rebuild of Mozilla Firefox)
• Tor Browser (a free and open-source project freely built upon the open source code of Mozilla Firefox)
• Opera (existed way before Chrome – Opera used to be one of my top browsers back then if I had to choose between Netscape Navigator and the good-riddance “Internet Explorer”. I am surprised myself that Opera still exists, and is currently owned and controlled by a Chinese company Kunlun Tech Co., Ltd.)
• Opera Air (World’s first browser with Mindfulness at its core ~ LOL!)
• Safari (for Apple users)
• Microsoft Edge (for Microsoft ecosystem users)

If you ask me: I am alternating between Brave and Vivaldi for various purposes. When I use Chrome is because I really have no choice since there are trails of past web developments.

In this entire list, you will see most of what Mozilla Firefox is powering from its open source codes made freely available to independent developers.

What AI Hallucination Actually is and Why It is Not a Bug

Most conversations about AI hallucination treat it as a malfunction. That framing is dangerously imprecise. Hallucination is not a flaw in the system. It is a predictable structural property of how large language models work, and understanding that distinction changes how a marketing team should approach AI-assisted content entirely.

An LLM does not retrieve information from a database of verified facts. It generates responses by predicting the most statistically probable sequence of tokens given an input. When you ask it for a statistic, it produces the kind of text that, in its training data, typically appeared alongside statistics: a plausible-sounding number, a source reference, and a vocally-confident sentence. Whether that number exists in reality is a separate question the model does not, and cannot, answer with certainty.

Anthropic’s model card documentation acknowledges this directly. The documentation describes current large language models as capable of producing outputs that are plausible-sounding but factually incorrect, and treats this as a known characteristic of the current model generation, not a defect awaiting correction. OpenAI’s system card documentation from the same period makes the same acknowledgement. These are not admissions of failure. They are transparent descriptions of how probability-based text generation works at scale.

For a consumer using an LLM to draft a personal email, hallucination is a minor inconvenience. For a B2B marketing team publishing AI-assisted content under a company byline in a market where editorial credibility is a primary trust signal, it is a categorically different category of risk. One that the production acceleration of AI tools makes structurally worse over time, not better.

The Brand Trust Stakes for B2B Marketers

In B2B commercial relationships, particularly across Singapore and the APAC region, trust precedes every purchasing decision. The buyer journey is longer, more risk-averse, and more dependent on the accumulated authority of the brands and individuals in a buyer’s professional orbit than in consumer markets. Editorial accuracy is not a content quality metric. It is a trust signal that prospects and clients evaluate, consciously or otherwise, with every piece of content you publish.

Consider what happens when a B2B brand publishes a white paper citing a Gartner statistic that does not appear in any Gartner report, or a practitioner commentary piece attributing a quote to a named industry expert who never said it. The first time a sophisticated buyer notices the discrepancy, they are unlikely to contact the brand to flag the error. They recalibrate their assessment of the brand’s editorial standards, quietly and permanently. In markets where referral relationships, long-term partnerships, and trust-first commercial dynamics drive revenue, that recalibration has consequences that are nearly impossible to reverse once they have occurred.

The AI content accuracy risk is structurally compounded by the production acceleration that AI enables. A content team that previously published four blog posts per month can, with AI assistance, produce 40. But a verification and editorial process designed for four posts per month does not automatically scale to 40. Hallucination risk increases in direct proportion to the productivity gains that AI-assisted content production is supposed to deliver. Without a deliberate verification infrastructure, scaling AI content production scales brand risk at the same rate.

RAG, Brand Context Layers, and the Architecture of Accurate AI Content

If your marketing team is deploying RAG-based systems for content production, research, or knowledge management, the single most important governance decision you will make is what goes into the context layer before retrieval begins.

RAG systems bridge the gap between large language models and an organisation’s knowledge corpus by retrieving verified, contextually relevant data at the moment of generation, ensuring AI outputs are both informed and trustworthy. Unlike generative AI powered by static pre-trained models, RAG grounds responses in real-time, curated, proprietary information.

The critical qualifier in that description is “curated.” Enterprise RAG fails without governance. Access controls, metadata, and context must precede retrieval. Context-graph-grounded RAG achieves up to five times improvements in AI analyst response accuracy over raw schemas. In plain marketing terms: the accuracy of what your AI content tool produces is only as good as the brand context you feed it before it retrieves anything.

For a B2B marketing team, a properly constructed brand context layer for a RAG-based content system should include five elements. Your canonical brand narrative and positioning statements, as approved and current. Your confirmed product and service specifications, including pricing, scope, and feature sets, with version dates. Your verified statistics library, where every figure has a source document attached. Your entity directory, covering named leadership, named clients, named partners, and named frameworks you use, each with accurate descriptors. And your prohibited claims list, the statements your brand has determined are not to be made under any circumstances, whether for legal, compliance, or accuracy reasons.

Without those five elements explicitly loaded into the context layer before your RAG system retrieves anything, the model defaults to its training data for anything outside its context window. And training data, by definition, includes everything the model was trained on, including inaccurate information about your brand, outdated product descriptions, and fabricated statistics that appeared on reputable-looking websites before anyone had verified them.

A Practitioner’s Note: Why I Am Not Fully Adopting Claude Design System Yet

I want to be transparent about where my Claude Design System test currently stands.

I have been experimenting with it as part of my own structured content and brand production workflow. My process started from a long time ago with TextEdit and Obsidian for quick information dumps, then moved into Markdown files, JSON files, and eventually a more organised design system structure. The goal was simple: reduce disjointed data, make my brand logic easier to reuse, and create a more efficient way to produce consistent outputs.

My experience so far? Useful, but not brilliant enough for full adoption.

Claude Design System works reasonably well for certain outputs, such as webpage concepts, slide decks, office suite formats, and broad visual direction. It can give you a fast starting point when you need something presentable and directionally aligned.

Where it becomes weaker is in granular brand execution. I am not convinced it is ready for building a proper UI kit, interface system, or brand design system that needs to preserve existing rules with precision.

The biggest issue is control. I do not want to spend too many tokens correcting the same instruction repeatedly, especially when session or weekly limits are involved. Being capped is not my favourite word when I need to process a string of tasks. If a system needs too much back-and-forth to understand basic design instructions, the productivity gain starts to disappear.

For example, when I say “follow the colour and interface design”, especially when I have attached a visual reference, I do not mean “recreate the whole UI kit” or override the design rules I already set. Those rules are not decorative preferences. They are part of my brand foundation.

At this stage, I am still unsure whether the issue is the Claude Design System itself, prompt interpretation, or the usual AI slop and buggy errors that appear when a tool tries to be too helpful. Either way, the gap is clear: it can generate outputs, but it does not yet understand my brand system deeply enough to protect it.

That is why I am not ready to recommend it as a complete end-to-end design system solution. It has value, but only within boundaries. For serious brand governance, interface consistency, and reusable design logic, I still need something more robust.

This is why I am using my own AI agents to help build a stronger design and brand system. Not to overcomplicate the process, but to create a workflow that can follow rules, preserve context, respect brand foundations, and scale across different product and content surfaces without constantly resetting the work.

The real lesson for B2B marketing leaders is this: AI design tools can accelerate production, but acceleration without governance is not a system. It is just output, and output is not the same thing as brand consistency.

This also connects directly to the wider issue of AI hallucination. Even capable AI tools require human brand context that they cannot generate for themselves. The tool does not know your voice, your verified claim library, your prohibited statements, or the standards your audience expects from you.

The tool is not the risk. Assuming the tool knows your brand as well as you do is the risk.

Singapore’s Agentic AI Governance Framework: What B2B Marketers Need to Know

On 22 January 2026, Singapore’s IMDA launched the Model AI Governance Framework for Agentic AI at the World Economic Forum in Davos, the world’s first governance framework specifically designed to address agentic AI systems. Released on 20 May 2026, the updated 51-page document incorporates extensive feedback from over 60 organisations including AWS, DBS, Google, Workday, OCBC, Tencent, and PwC. The update adds guidance on risks linked to multi-agent systems, third-party agents, automation bias, and human accountability.

The framework’s core accountability principle is unambiguous: compliance is voluntary, but organisations remain legally accountable for their agents’ behaviours and actions. For B2B marketing teams in Singapore deploying AI in content production, this means two things in practice. First, the brand is the publisher of record, regardless of which tool generated the first draft. Second, human accountability for AI outputs is not optional or aspirational. It is the framework’s baseline expectation.

The MGF applies to all organisations deploying agentic AI in Singapore, whether or not they are training or developing in-house AI agents, or using third-party agents. If your team is using a third-party AI writing tool, a RAG-based content system, or an agentic workflow that produces customer-facing outputs, the framework’s human accountability requirements apply to you. For B2B brands working with clients in regulated industries such as financial services, healthcare, or professional services, this governance context is increasingly a commercial consideration. Documented AI content governance processes are becoming a differentiator in agency and consultancy selection conversations.

A Three-Step Verification Framework for AI-Assisted Content

The practical question for B2B marketing teams is not whether to use AI in content production. Most teams are already using it, and the productivity case is well-established. The question is how to introduce a verification layer that reliably catches hallucinated claims before they reach a published post, a distributed white paper, or a client-facing report.

Step 1: The Source Audit

For every statistic, percentage figure, research finding, or authoritative claim in an AI-generated draft, locate the original source document and verify that the claim appears in that document in the form stated. Not a paraphrase that subtly shifts the meaning. The exact claim, in context. If the original source cannot be located within five minutes of active searching, the claim must be rewritten as practitioner opinion clearly framed as such, or removed from the draft entirely.

This step sounds straightforward, but most teams skip it because AI output sounds authoritative, and authoritative-sounding text is psychologically difficult to challenge without a systematic process that requires you to do so regardless of how confident the sentence reads.

Step 2: The Entity Review

Examine every named company, named individual, named framework, product, or initiative that appears in the draft. Confirm that each entity exists, that it is described accurately in the context it appears, and that any quoted statements are correctly attributed. Large language models are particularly susceptible to entity-level errors: attributing real statements to real people who never made them, or describing real companies as having launched products that belong to entirely different organisations. An entity review takes three to five minutes on a standard blog post and catches a disproportionate share of the hallucinations that create the highest brand risk with the most sophisticated readers.

Step 3: The Logic Audit

Read the draft’s conclusions against the evidence it cites and ask whether the conclusion actually follows from that evidence. AI-generated content frequently produces logical non-sequiturs: citing a study on consumer behaviour to support a claim about B2B purchasing patterns, or applying a regional statistic from one market to support a global generalisation. This is not always hallucination in the technical sense, but it produces the same practical problem: a claim that cannot withstand scrutiny from a careful, knowledgeable reader.

Applied consistently, this three-step process takes between 15 and 30 minutes for a standard blog post. If a team’s production volume makes systematic verification time prohibitive, the correct operational response is to reduce AI output volume to what can be verified to a professional editorial standard, not to skip verification and accept the cumulative brand risk that follows.

The Verification Advantage in a Zero-Click World

In a search landscape where 93 percent of AI Mode queries end without a click, the content that earns citations in AI-generated answers is the only content doing brand-building work at scale. RAG is the filter that determines which brands appear in AI-generated answers. If your content is not retrieved at the retrieval stage, your brand cannot be cited, recommended, or even mentioned, regardless of how much content you publish or how much you spend on traditional SEO.

The brands that will build durable authority in this environment are not those that produce the most AI-assisted content. They are those that produce AI-assisted content that is accurate, source-backed, and editorially defensible. In markets where trust precedes every commercial decision, that distinction compounds over time. A reputation for editorial rigour, built through consistent verification practice, becomes a competitive advantage that is difficult to replicate at speed.

If you are leading a B2B marketing function in Singapore or the broader APAC region and you do not yet have a documented AI content verification process, building one is the highest-return governance action available to you this quarter. The three-step framework above is a practical starting point, designed to be repeatable without requiring specialist AI expertise. Apply it consistently, refine it as your team’s production workflow evolves, and treat it as a non-negotiable editorial standard.

My LITV AI SEO Agent 2.0 audits your AI search visibility across ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini, and Bing Copilot, identifying where AI systems are generating incorrect or absent information about your brand and providing a structured fix pack for closing those gaps. Start your free access and find out what AI search is currently saying about you.

WordPress powers 42 percent of the Internet (nearly half the web) and right now, that number is moving in the wrong direction for the first time in over a decade. 

My Personal Anecdote (My WordPress Journey, and Why the History Matters)

When I launched LadyinTechverse in early 2025, I chose to build it on self-hosted WordPress over a DigitalOcean droplet with almost no server-side experience because I wanted to own every layer of what I was building, the code, the database, the content, and the direction, which is opposite of the easiest path. WordPress was not just a platform choice. It was a declaration that this digital home belongs to me, and that no SaaS vendor decision, pricing change, or product pivot could take it from me. 23 years after its first release, that founding principle is still the most compelling reason to stay. The GPL licence that underpins WordPress made all of that equally accessible to all of us, and that is why the foundation still exists. 

That personal decision did not happen in a vacuum. I was exposed to early WordPress development in the 2000s, when the web was in mid-transition from Web 1.0 to Web 2.0, and the question of who could publish on the Internet was genuinely open for the first time. WordPress was released on 27 May 2003, forked from b2/cafelog by Matt Mullenweg and Mike Little, while Drupal had been available since 2000, and Joomla arrived in 2005. Together they offered something radical – free, flexible, and community-governed infrastructure for anyone who wanted to publish, build, or sell online without asking permission from a platform that owned their content. Built-in WYSIWYG text editing, SEO, upgraded HTML rendering, and a graphical user interface that required minimal coding arrived as a full content publishing package, covering images, files, pages, databases, and texts. Squarespace launched in the same year as WordPress in 2003, with Weebly following in 2006, opening the no-code templated end of the market to people with no knowledge of HTML or CSS at all. On the monetisation side, Google acquired Blogger in February 2003 and made it freely available to the world, pushing the entire concept of blogging into mainstream culture. Thereafter, Google AdSense launched publicly on 18 June 2003, handing bloggers without large platforms a genuine path to earning from their writing for the first time. The democratisation of publishing and the democratisation of monetisation arrived together in the same calendar year, and WordPress sat at the intersection of both. That is the historical context inside which this platform’s 23 years of relevance should be read and commemorated.

Six Months of Decline and the Release that Arrived to Boost

What the Numbers Say

WordPress has recorded six consecutive months of declining market share, dropping from 43.2 percent in December 2025 to 41.9 percent by May 2026. That 1.1 percentage point fall represents roughly four times the year-over-year decline rate measured between January 2025 and January 2026. W3Techs data is corroborated by HTTPArchive adoption metrics showing identical directional movement, which means this is not a measurement artefact. It is a real trend.

What makes it meaningful rather than just statistical data is that it is largely a WordPress-specific issue. Virtually all other major content management platforms are holding steady or growing in the same period. Shopify, Wix, and Squarespace are all gaining modest ground. Wix recorded 32.6 percent year-over-year growth. The market is not contracting. WordPress is losing share to competitors who have been sharpening their propositions against it for years.

What arrived at exactly this moment of scrutiny is worth noting carefully. WordPress 7.0, named Armstrong, launched on 20 May 2026 as the most structurally significant release since Gutenberg shipped in 2018. Native AI integration sits at the centre through the WP AI Client and Abilities API, a provider-agnostic framework connecting directly to OpenAI, Anthropic’s Claude, and Google Gemini without requiring a separate plugin for each integration. More than 400 enhancements and 486 bug fixes shipped across the editor, dashboard, and AI integration layers. How that AI layer connects to the broader shift in how content gets discovered is explored in the LITV analysis of how AI search is reshaping organic visibility.

A platform building that kind of infrastructure at the exact moment its numbers are being questioned is not winding down. It is recalibrating, and the difference between those two things matters enormously for how you make your next technology decision.

The Part Nobody Wants to Name Directly

Why the Governance Question Will Not Go Away

The market share decline is not only a technology story. It is a trust story, and honest analysis requires naming what actually happened rather than softening it into vague industry commentary.

Automattic’s valuation peaked near 7.5 billion dollars following a 2021 funding round. Since then, institutional investor BlackRock has marked down its stake by approximately 63 percent, with shares valued at 31.03 dollars as of March 2025 against an original purchase price of 85 dollars per share. Automattic laid off 16 percent of its global workforce in April 2025, roughly 281 employees across 90 countries. The legal dispute with WP Engine, triggered when CEO Matt Mullenweg blocked WP Engine from accessing WordPress.org, produced a disclosure that genuinely alarmed the developer community. Legal filings claimed that WordPress.org is Matt Mullenweg’s personal property rather than an asset held in trust by the WordPress Foundation.

For hundreds of thousands of businesses and developers who built their digital infrastructure on the assumption that WordPress is an open-source commons governed by a neutral foundation, that is not a minor clarification. It is the kind of governance uncertainty that makes enterprise architects nervous and individual developers quietly start evaluating their options. The question of why strategy and ownership clarity matter more than any individual tool sits underneath all of this.

I am writing this so as not to keep piling on about politics going on within the WordPress community itself. The full picture is having the only honest basket of eggs, and aside from the community (Business owners, enterprise users, developers, independent solo operators, professionals, hobbyists and students) deserves the full picture.

What 42 Percent Shared of the Internet Means

The Ownership Argument that SaaS Cannot Win

WordPress still holds 59.9 percent CMS market share among sites using a known content management system, and powers an estimated 472 to 595 million websites globally. The organisations running on it include Accenture, IBM, Samsung, Shell, JPMorgan Chase, and the United Nations, deploying it for blogs, career portals, enterprise content layers, and microsites at institutional scale. These are not casual or one-off technology decisions made by teams who had not considered the alternatives. They are deliberate choices made by people who understand exactly what they are buying.

The reason enterprises and solopreneurs alike keep choosing self-hosted WordPress comes down to something no SaaS website builder can replicate: full code ownership, full customisation depth, self-hosting capability, and the freedom to extend the platform in any direction without asking permission from a vendor’s product roadmap. WooCommerce processed an estimated 35 billion dollars in gross merchandise volume in 2025, making it the largest open-source eCommerce platform by transaction volume globally.

The SaaS builders gaining ground are genuinely capable tools for specific use cases. But they operate on a fundamentally different ownership model. You are renting space on someone else’s infrastructure, governed by someone else’s terms, constrained by what their feature decisions allow. The day that vendor changes pricing, deprecates an integration, or rewrites API access policy, your digital presence is at their discretion. Self-hosted WordPress is yours forever, like a diamond. The code. The database. The content architecture. The migration path. That ownership matters deeply to anyone building something intended to outlast any SaaS website platform or third-party vendor’s current strategic priorities, and the rise of vibe coding and low-code alternatives makes the ownership conversation more urgent rather than less.

What the Community Can Do Right Now

Keep Building on It

WordPress 7.0’s native AI Client with pre-registered connectors for OpenAI, Anthropic, and Google introduces a standardised AI layer that the 70,000-plus plugin developers in the ecosystem will build on. The integrations extending the platform’s relevance into workflows that barely existed two years ago are coming, and they will arrive faster because the infrastructure is now standardised in core rather than fragmented across competing plugin approaches. Building on the platform right now is a direct vote for its continued momentum.

Pressure the Governance Conversation

The questions raised by the WP Engine dispute about WordPress.org’s ownership and the WordPress Foundation’s actual authority deserve continuous serious community attention. These should not be settled in a corporate lawsuit and quietly forgotten. The governance structures protecting the open-source commons belong to the community that depends on them, not to any single executive or company. Asking those questions loudly, consistently, and with specificity is the most useful thing a community member can do right now.

Invest in Performance

The average WordPress page loads in 3.4 seconds against the 2.5-second threshold Google recommends for Core Web Vitals. Modern alternatives like Next.js average 0.8 seconds, and that gap is cited as a migration reason by departing developers with increasing regularity. This is a real competitive vulnerability and it is a solvable one. Managed hosting, proper caching configuration, image optimisation, and the client-side media processing built into WordPress 7.0 collectively address the speed gap for teams willing to treat it as an implementation discipline problem rather than a platform limitation.

Contribute, Even in Small Ways

The open-source commons that makes WordPress available to a Malaysian developer building their first client site, a solopreneur in Singapore scaling a personal brand, and a Fortune 500 enterprise architecting a global content system is sustained by human contribution. WordCamp. GitHub contributions. Plugin reviews. Forum answers. Documentation improvements. Every person who participates in that ecosystem is casting a vote for its continued relevance, and those votes compound in ways that no marketing campaign can replicate.

My WordPress Story, and Why I Still Love WordPress

I run ladyintechverse.com on WordPress.org. The platform hosting my writing, my brand, my products, and more than a year of content I own completely – I can take anywhere if I ever need to. WPCode handles my custom scripts. A custom theme reflects my LITV brand system. My posts sit in a database I control on the hosting infrastructure I chose. That ownership matters to me as a founder building a long-term brand, and it will matter to you the day a SaaS platform you trusted decides to change its pricing model, deprecate the feature your workflow depends on, or simply decide that your use case no longer fits their product roadmap.

None of this means the data should be ignored, the governance questions should be left unanswered, or the performance gap should be accepted as a permanent condition. All three deserve serious attention from anyone who cares about where this platform goes next. Trust, once built through transparency and consistency, is the hardest thing in technology to rebuild when it fractures, and how brands sustain trust through genuine integrity rather than performance is the question sitting underneath this entire conversation.

Series: Building in Public Journal | Week 4 of 4

There is a difference between a tool and a product. A tool runs once. A product has to keep working when a user signs in, runs an audit, reads the result, asks what to fix, exports the work, comes back next week, and gives feedback when something feels off. Week 4 was the week LITV AI SEO Agent 2.0 crossed that line. That is the distinction that defined Week 4. And it is the most honest thing I can say about what this sprint taught me.

If you are joining this series for the first time: I have been building the LITV AI SEO Agent 2.0 live during a four-week sprint called the Women in AI Accelerator, run by Build Club and hosted by Annie Liao and Caroline Ciaramitaro. Week 1 was architecture and choosing not to build the easy version. Week 2 was auth, staging, and the first real code commit. Week 3 was the unglamorous compliance and layers of backend work that holds the entire structure together. Week 4 was where the loop had to stick and the product around the engine had to materialise.

From Engine to Product: The Shift that Defines What Shipped in Week 4

The audit engine was always the centrepiece of the build. It crawls publicly accessible URLs, analyses them across Technical SEO, SXO, GEO, and AEO signal layers, and returns a more structured intelligence report. It runs in a loop: it enters a website, processes the allowed pages, aggregates the scoring, delivers the output, and waits for the next request.

What Week 4 made clear is that the engine being operational is not the same as the product being usable. The harder work is everything around the output: access control, delivery, data handling, error states, mobile behaviour, admin workflows, and the feedback path that tells you what users actually experience when the engine does its job.

That surrounding layer is what Week 4 was for. Here are the five highlights that defined it.

5 Key Highlights from the Week 4 Build

1. The Audit Loop Held and What that Means

This is the technical win I am most proud of from the entire sprint. The audit loop has to function at three stages: initial entry into the target URL, mid-stage signal processing across the four audit frameworks, and final output generation into a structured report from the layers of audit and visibility. If it breaks at any one of those stages, the failure is clear as day. If it holds across all three, I have the evidence that the architecture is sound.

The loop broke three times during Week 3 before it held consistently. Two breaks were in the signal extraction layer, where edge cases in structured data markup were not being handled correctly. Each break was fixed directly in the engine’s core logic, then patched through to the surface. The loop now holds on standard sites and on slow-crawl days where the pipeline is under more stress and on downtime days.

An AI-powered engine that fails silently is not an engine. It is a liability that breaks the loop and can run simple autofixes.

2. The Dashboard became Usable, unlike a Scratch Surface

The dashboard had to stop being fragile. Week 4 addressed the layout across desktop, tablet, and mobile views — tightening panels, making wide issue tables scroll inside their own containers, and making the interface feel calmer and more consistent across different screen sizes.

Accessibility hardening was also part of this. This is not the glamorous part of an AI product build. But it is the part users feel immediately, and the part that determines whether someone who is not technical trusts the product enough to use it a second time.

3. The Fix Pack became a part of the Workflow, Not as a Download

The Fix Pack is the action layer of the audit. It takes the findings from the Technical SEO, SXO, GEO, and AEO report and turns them into a prioritised set of recommendations that a user can actually work through.

In version 1.0, Fix Pack output was just a static download. Week 4 moved it towards a proper execution workflow. Paid users can now work with PDF and CSV outputs through clearer delivery paths, and the product is materialising to support the rhythm I care about most: audit, identify the gaps, prioritise the fixes, validate the key blockers, and re-audit for better indexing to take place. That cycle is the actual value proposition of the product. A dashboard full of scores with no path forward is just anxiety in a UI.

4. Telegram Delivery Got a Proper Verification Flow

Telegram delivery has been on the roadmap since early in the build. The idea is straightforward: users who want their audit results and Fix Pack outputs delivered directly to Telegram can connect the LITV bot, @litv_agentic_bot, through a clean verification step rather than manually pasting chat IDs or guessing what to do inside the bot. Week 4 tightened the backend and frontend around that flow. This matters for non-technical users who would otherwise abandon the setup at the first point of ambiguity.

5. Beta Users will Soon Be Inside the Loop

Beta licence keys are prepared for Build Club testers to access the paid-tier dashboard experience during the upcoming testing window. A standalone beta feedback form was built intentionally separate from the main product navigation to capture product feedback.

Final Conclusion: Nothing is Final When You are Building Something Serious

The Women in AI Accelerator is a four-week sprint, and the LITV AI SEO Agent 2.0 is not a four-week product. The sprint gave me structure, community support, accountability, and the enforced function to ship what I had been planning since before I joined the accelerator. But nothing about this build is 100% finished because nothing about the problem it solves is finished.

Why is it Not Finished?

AI search is evolving. The platforms that mediate it are evolving. The signals that determine citation and visibility will shift, affecting some of them before the end of this year. A product built to audit those signals has to be built with the assumption that the audit criteria will change. That is not a caveat. It is the full product and design brief.

The comprehensive audit and visibility engine inside LITV AI SEO Agent 2.0 was always the centrepiece of the build. But an engine without a usable product around it is infrastructure, and not a service. Week 4 was where the two finally met and they clicked.

What’s Coming for Demo Day

TBC.

Thanks to Build Club, Annie Liao and Caroline Ciaramitaro for hosting and facilitating this 4-week immersive AI building programme.

Series: Building in Public Journal | Week 3 of 4

There is a particular kind of discipline required to build boring things on purpose. Not that I ran out of ideas, not because the sprint ran out of steam, but because I made a deliberate choice to prioritise what holds the entire structure together over what looks impressive in a demo.

That is what Week 3 was.

If you are joining this series for the first time: I am building the LITV AI SEO Agent 2.0 live during the four-week Women in AI Accelerator run by Build Club. The sprint is a structured programme that pushes you to ship a meaningful product in a compressed timeline in public, and with accountability deposited.

Week 1 was architecture. Week 2 was auth and the first real code commit. You can read both in the Week 1 build journal and the Week 2 build journal. Week 3 was the work nobody photographs.

Why Compliance Week is Not a Consolation Prize

Most build-in-public journals skip this part. They go straight from “auth is live” to “dashboard is beautiful.” The compliance layer sits quietly in between, undocumented and unappreciated until something breaks and suddenly the business is exposed.

I built Week 3 the way I did because I have spent enough time in marketing and communications to know that trust is not earned through features. It is earned through consistency, transparency, and the things users see before they ever click a button. A privacy policy is not a legal formality. 

For the LITV AI SEO Agent 2.0, which audits websites for AI search visibility and generative engine optimisation, credibility is the product. If I am building a tool that helps brands earn trust in AI-mediated search, that tool has to demonstrate exactly the same behaviour it advises.

What Shipped in Week 3

1. Weights Rebalanced and Disclosure Page Published

The AI Visibility Index is one of the core scoring engines inside the LITV AI SEO Agent 2.0. It measures how visible a brand is across the major AI-powered search platforms: ChatGPT, Google AI Overviews, Perplexity, Gemini, Bing Copilot, and Claude.

These weightings reflect current market share and citation frequency data, not arbitrary preference. They will be updated as the landscape shifts.

Every user who receives a specific audit score can see exactly how that score is calculated, which platforms are included, what the methodology is, and when it was last updated. AI search is a citation game. If this tool asks brands to be transparent and citable, the tool itself has to model closely to that behaviour first.

2. Consent and Analytics, Done Properly

Analytics and audits are wired in, but not in the way that fires everything the moment someone lands on the page. A consent gate is in place. EU and UK visitors see a consent banner before any tracking fires. 

This matters more than most product builders acknowledge. Consent-less analytics is not just a legal risk; it is a data quality problem. Traffic data collected without user consent produces unreliable numbers that lead to poor decisions. Building the consent layer properly from the start means the data that does come through is not only clean but trustworthy.

3. Legal Pages Live

Privacy policy, terms of service, cookie policy, and contact details are all live and styled to the light and dark themes. These were not templated and left to gather dust. Each section was reviewed against the actual data flows in the product. What is collected, where it lives, how long it is retained, and who can request deletion. It took longer than expected, and it always does.

4. Single Pro Tier

The Starter tier is gone. There is now one product plan and that is it.

The decision to kill Starter was not spontaneous. The original logic behind a Starter tier was to provide a low-commitment entry point. What it actually created was decision fatigue at the pricing page, a more complex licence webhook to maintain, and a middle-tier ambiguity problem where users could not clearly identify what they were getting for the price differential.

One tier eliminates all of that. The free access level remains for users who want to run a single audit and see their scores. Pro unlocks the full access without upselling the dashboard workflow, full audit, AI Visibility tools and more. The value distinction is clear. The pricing page is now a single decision.

5. Crawler Hardening

On the backend, the crawler that powers the technical audit pipeline received several meaningful updates. The maximum pages per audit is capped. A minimum two-second delay between requests is enforced to ensure the agent is not acting as a hostile scraper against the sites it is auditing. A bug in the robots.txt parser was fixed. Multiple backend tests are passed. The user agent string is now declared as LITV-AISEOAgent/2.0, which is a small but important commitment: the agent identifies itself honestly when it crawls.

Why the Boring Week is the Most “Sweatshop” Week

Here is the honest CMO take on Week 3. Compliance and pricing clarity are not housekeeping tasks you squeeze in before the exciting work. They are the foundation on which every acquisition and retention decision rests.

AI search is a citation game, not a ranking game. I have written about this in the context of AI Overviews reducing organic clicks. The same logic applies here. Brands that want to be cited by ChatGPT, Perplexity, and Google AI Overviews need to be structured, transparent, and machine-legible. A tool that audits for those qualities has to demonstrate them itself. Publishing the methodology is not a nice-to-have as it is a launchpad for orchestrating and architecture decision-making.

Every one of these choices is a signal to the user about what kind of product this is and what kind of builder is behind it.

What is Coming in Week 4

Week 4 is the polished dashboard that is the centre of the universe here. With one-click re-audit from the user’s view, and the AI Visibility Tools that let users see where their brand is and is not being cited across the six platforms tracked by other specifications that I am still considering to add on.Week 3 was the foundation of the scaffolding work. Week 4 is what cleaning up means. In between Claude Code, ChatGPT, LLM models, ChatGPT 5.5 became my sweeper who would occasionally flagged dirty worktrees and the time is up to clean. It is like similarly how I would treat my spreadsheets. To be honest, I cannot just rely on one AI model like Claude Code to handle everything. I’d still need to keep an open mind to allow other models to contribute and that’d eventually lead me to better decision-making. In fact, due to the heavy traffic uses and certain peak hours of Claude Code and Claude, the AI slops can generally occur, and there are occasions when I had to stop using it.

Thanks to Build Club, Annie Liao and Caroline Ciaramitaro for hosting and facilitating this 4-week immersive AI building programme.

Series: Building in Public Journal | Week 2 of 4

Week 1 was architecture on paper. Week 2 is where the build gets real.

If you missed the first entry in this series, the short version is this: I am four weeks into the Women in AI Accelerator run by Build Club, and I am using the sprint to upgrade the LITV AI SEO Agent 1.0 from a single-purpose audit tool at version 1.0 to a full system engine at 2.0. Week 1 was about locking the architecture and choosing the harder future-proof build path over the faster and shakier one. You can read that full decision log in the Week 1 build journal. Week 2 was about shipping.

What Week 2 was actually About

The Week 1 decision to build properly with authenticated product layers, real plan tier logic, and a migration path for existing users meant that Week 2 had a very clear mandate: get the auth layer live on staging. Literally live, working, tested.

That sounds like one task. I kid you not. Authentication is one of the most consequential infrastructure decisions in any application or a SaaS product, and getting it wrong early creates the kind of technical debt that follows you into every subsequent sprint. I had already argued this point in Week 1 and lost to the better argument. Week 2 was the proof of that choice.

The Key Decision that Shaped the Entire Week

The single most important call I made in Week 2 was choosing Supabase Auth over a license-key-only model. The license-key approach would have been faster to ship.

Supabase Auth, specifically the auth flow, captures a verified email from the first interaction, requires no password, and eliminates the single biggest friction point at sign-up. The upfront cost was an additional four to six hours of build time in Week 2.

What Shipped in Week 2

24 files shipped across the API backend and the React frontend. Here is what that means in plain English without the full technical breakdown.

Passwordless authentication is live. Users log in via a secure link sent to their email with no password required. The API is protected with tokens and three permission levels. A rate limiter is in place for keeping the product layer sustainable without open abuse. The dashboard shows each user their current tier, their remaining audit quota, and their last few remaining audits. And critically, existing 1.0 users have a clean migration path to bind their license keys to the new account system in a single step, with no loss of access.

The Answer Engine Optimisation (AEO) layer was also significantly expanded this week. For context, AEO is about making your content legible and citable to AI-powered answer tools like ChatGPT, Perplexity, and Google AI Overviews. As I wrote in my earlier post on how AI overviews are reducing organic clicks, the brands that are not structuring their content for AI-mediated search are losing visibility they cannot easily recover. The expanded AEO layer now checks for specific schemas, content signals, complete structured data coverage, and a portion of tech coding validity, among others. These are not edge cases as they are now primary signals in how AI search tools evaluate and surface content.

What Building Solo actually Feels like at this Stage

There is a particular kind of pressure that comes with building a product alone on a four-week sprint while that sprint is publicly tracked. Week 2 had that pressure in full. The auth layer works but still need a bit more of the “hit the nail on the head”. The migration path works. But there were also three separate moments this week where something did not work, required a lot of debugging at a level that consumed hours I had not planned for, and sent me back to the architecture with fresh questions I had not anticipated.

This is the part of building in public that I think is genuinely underrepresented in the content most builders share. The progress posts tend to celebrate the wins and skim past the hours of confusion, the dead ends, a bit of AI slop at times, and the moments where the right answer only becomes obvious after you have tried a few wrong turns. I am not skimming past those here – they did happened and in turn, my glucose level dropped immensely. These factors make up a part of the build, and they did not stop the delivery since they can be costly for solo builders who should not be surprised at all.

The build is on schedule. Week 3 has started.

What is Coming in Week 3

Week 3 is the deep abyss of GEO engine rewrite, which has been the most strategically anticipated component of the entire 2.0 system since I locked the GEO Audit Engine Spec v3.0 in Week 1. The GEO engine is where the LITV AI SEO Agent 1.0 is quite the most differentiated from the tools already available in the market. Some SEO audit tools treat GEO as an afterthought. This build treats it as a primary discipline with its own signal architecture and output framework.

I will document the rewrite as it progresses and publish the Week 3 update here.

The current version of the agent, LITV AI SEO Agent 1.0, is live and free to use at seoagent.ladyintechverse.com. Run an audit. See exactly what 2.0 is being built to surpass. And check back here for Week 3.

Thanks to Build Club, Annie Liao and Caroline Ciaramitaro for hosting and facilitating this 4-week immersive AI building programme.

Series: Building in Public Journal | Week 1 of 4

I joined a room of 140+ women building with AI, and the first thing I did was decide to make my own product significantly harder to build. That probably sounds counterproductive. But if you have ever been at a genuine build fork, where the easy path and the right path are not the same road, you already know why I made that call.

This is Week 1 of my building-in-public journal, tracking the upgrade of the LITV AI SEO Agent from version 1.0 to a full system engine at 2.0. I am writing this as a practitioner log, not a polished case study. That means the decisions, the pivots, the arguments with my AI collaborator, and the architecture choices are all going raw in here. Welcome to the build, and a bumpy momentum indeed.

Why I Joined the Women in AI Accelerator by Build Club

The Women in AI Accelerator is run by Build Club, an Australian-founded community supporting women building with AI tools. The kick-off brought together over 150 builders for a four-week sprint, each working on their own AI product under a shared challenge brief. You can find Build Club at buildclub.ai and follow the programme on LinkedIn. The community is exactly what it sounds like: real builders with humility, actual progress tracked in public.

For me, joining was a deliberate strategic decision. I had been running the LITV AI SEO Agent 1.0 at seoagent.ladyintechverse.com as a live freemium product since early March this year, offering Technical SEO, SXO, GEO and AEO audits to marketers, founders, and solopreneurs who want actionable visibility intelligence without a bloated MarTech stack. The product works. But version 1.0 was built as a proof-of-concept, not ready yet to scale the system. The Women in AI Accelerator gave me a structured four-week container to do what I had been planning: rebuild the engine properly.

What I did not expect was that Week 1 would force me to make one of the clearest product architecture decisions I have faced since launching LadyinTechverse. More on that shortly.

What the LITV AI SEO Agent 2.0 Upgrade Means

Version 1.0 of the LITV AI SEO Agent operates as a single-purpose audit tool. You enter a URL, it runs the audit pipeline across Technical SEO, SXO, GEO and AEO signals, and you get a structured, actionable report. It does that well. The free tier allows multiple audits per day without a login requirement, which keeps the crawling friction low and the entry point open. 

Version 2.0 is a different proposition altogether. The goal is to move from a single audit tool to an AI-powered SEO, visibility and content system engine, layered with an audit engine that is constantly checking and tracking search queries. That means combining Technical SEO, SXO, GEO and AEO audits into a single intelligence layer, adding authenticated product tiers with differentiated access, integrating richer data connectors, and building the GEO audit engine from scratch using a more robust specification. The technology choices are not accidental. This is not just a feature update. It is a product foundation rebuild with a longer commercial runway in mind.

The Technical Specification v1.1 and GEO Audit Engine Spec v3.0

In the early days of this first week, I finalised the Technical Specification at version 1.1 and locked the GEO Audit Engine Specification at version 3.0. For those unfamiliar with what a GEO audit engine does: GEO stands for Generative Engine Optimisation, the practice of making your content legible and citable to large language models such as Claude, ChatGPT, Gemini, and Perplexity. As AI search continues to reshape how users discover information, GEO is no longer optional for brands that want to maintain visibility in an era where AI overviews are intercepting traffic before it reaches your site. I wrote about this directly in my post on AI overviews and brand visibility, which is worth reading alongside this build log.

Locking the GEO Audit Engine Spec at v3.0 was not a trivial milestone. It represents a significant upgrade in the signal categories the engine will evaluate, the depth of analysis per category, and the output format that makes the audit genuinely actionable rather than just informative. The full architecture is confirmed, and the mix of backend coding languages is locked in. That is the foundation you build everything else on, and getting it right in Week 1 saved me from a much more painful rearchitecture later.

The Fork in the Build: Easy Path vs the Right Path

Here is the part of Week 1 that I was not planning to write about publicly, but it is probably the most useful thing in this entire post. Midway through the week, I hit a genuine product architecture fork. The choice was between two build approaches: the easy build and the future-proof build.

The easy build gets you to a working product faster. You make pragmatic tradeoffs in the infrastructure, you skip some of the harder authentication and product layer decisions, and you ship something that functions well enough for the current user base. Some builders would probably choose this because time is of the essence. The time pressure is real, the market moves fast, and a working product beats a perfect one on paper.

The future-proof build is much harder. It requires setting up proper authenticated product layers from the beginning, which takes longer in Week 1 but prevents you from having to gut the product architecture later when scale or complexity demands it. It is the kind of decision that costs you this week and saves you six months in 12 months’ time.

I debated and argued to get the decision out with Claude, my primary AI collaborator, as well as ChatGPT, my systems collaborator for this build. I laid out both paths, the tradeoffs, the timelines, the downstream implications. And Claude, made the better argument for the harder build. I pushed back. I lost the argument. And I am glad I did, because the future-proof build is unambiguously the correct call for a product that is intended to eventually merge into an envisioned future ecosystem. The authenticated product layers are now in place, and the build is on schedule. Crossing fingers.

If you are building something serious, read that last paragraph again. The easy build is almost never the right build when you have a longer personal usage or commercial horizon. The time you think you are saving in Week 1 becomes a technical debt that compounds faster than you expect.

What Week 1 Delivered: Architecture Locked, GEO Engine Confirmed

At the close of Week 1, the full product architecture is validated against the target coding language mix for the backend. The authenticated product layers that will support the 2.0 build are structured and ready. 

That might sound like a lot of paperwork and not much coding. But anyone who has ever shipped a full stack production of an AI product will tell you that the architecture decisions made in Week 1 are the decisions you live with for the lifetime of the product. Getting them right is not administrative work. It is the most consequential engineering work in the entire build.

Week 2 begins with the first code commit: rewriting the GEO engine from scratch, based on the locked specification. That is where the build gets visibly messy and haywire. If you are curious about where the LITV AI SEO Agent currently stands and what a free audit feels like, you can run one at seoagent.ladyintechverse.com. The 1.0 engine is live, and it gives you a clear sense of what 2.0 is being built to surpass.

What is Coming in Week 2

The first code commit lands at the start of Week 2, and the focus is singular: the GEO engine rewrite. This is the component I consider the most strategically differentiated part of the entire 2.0 system. Most SEO tools still treat GEO as a footnote or a bolt-on checklist. The LITV AI SEO Agent is being built to treat it as a primary audit discipline, with its own specification, its own signal architecture, and its own output framework.

I will document week 2 in the coming days, and share the key decisions here in the Week 2 update. If you want to follow the build in real time, the best place to do that is here.

Building in public is not a marketing strategy for me. It is a commitment to transparency about what AI product development actually looks like when you are doing it solo, without a team, without external funding, and with uncharted terrains ahead that goes beyond the four-week sprint. Week 1 is done. Week 2 starts now. Run your free SEO audit at seoagent.ladyintechverse.com and follow the build here.

Thanks to Build Club, Annie Liao and Caroline Ciaramitaro for hosting and facilitating this 4-week immersive AI building programme.