Which GEO Tactics Has Google Officially Said You Don't Need? The May 2026 Guide, Point by Point

Google's central statement is worth quoting exactly, because it settles a debate that ran for two years:

"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."

The technical reason follows: Google's generative AI features are rooted in its core Search ranking and quality systems. The AI layer sits on top of the index, not beside it. The same index and quality signals that earn rankings are what earn citations in AI answers.

Much of the trade press read this as Google deflating the GEO category. There is a more accurate reading: on 5 June 2026, Google separately updated its "Do you need an SEO?" hiring guide to name AEO and GEO as legitimate services a credible SEO can provide — the first time that foundational document has done so. Google did not say GEO is fake. It placed GEO inside a discipline that already has standards and a public quality bar. That is legitimisation with guardrails, not dismissal.

Six: llms.txt and other AI-specific files, content chunking, rewriting content specifically for AI, special AI schema, manufactured brand mentions, and any vendor claim of being "approved by Google." On 15 May 2026, Google published its first official guidance on optimising for generative AI features — "Optimizing your website for generative AI features on Google Search" — and its most quoted section is titled, bluntly, "what you don't need to do." Google names these tactics directly, most with a verbatim "you don't need to" statement. Four are declared unnecessary; two are actively warned against. For any Singapore business currently being sold a GEO package, this document is the most useful procurement tool published all year, because it lets you check every line of a proposal against Google's own words. This article maps each tactic against the official position.

One scoping note before anything else, because most coverage blurs it: this guidance applies to Google's AI features only — AI Overviews and AI Mode. ChatGPT, Claude and Perplexity run on different indexing and retrieval models and may respond to things Google dismisses. Any vendor treating "AI optimisation" as one undifferentiated thing across all platforms is over-generalising.

The Framing: "Still SEO"

Google's central statement is worth quoting exactly, because it settles a debate that ran for two years:

"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."

The technical reason follows: Google's generative AI features are rooted in its core Search ranking and quality systems. The AI layer sits on top of the index, not beside it. The same index and quality signals that earn rankings are what earn citations in AI answers.

Much of the trade press read this as Google deflating the GEO category. There is a more accurate reading: on 5 June 2026, Google separately updated its "Do you need an SEO?" hiring guide to name AEO and GEO as legitimate services a credible SEO can provide — the first time that foundational document has done so. Google did not say GEO is fake. It placed GEO inside a discipline that already has standards and a public quality bar. That is legitimisation with guardrails, not dismissal.

The Four Tactics Google Says Are Unnecessary

1. llms.txt and AI-specific files

The pitch: place a machine-readable manifest at your site root so AI systems can find and prioritise your content. Some platforms charge a monthly fee for "GEO file generation."

Google's position, verbatim: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search."

Google may crawl an llms.txt file as it would any page, but indexing a file does not mean it receives special treatment. For Google's AI features specifically, that monthly fee buys nothing.

The honest nuance: llms.txt is not dead everywhere — there is a more credible case for it with AI agents and documentation tooling, and adoption is growing in those contexts. But for the largest AI search surface in the world, Google has now said plainly that it does nothing.

2. Content chunking

The pitch: break long content into tiny 50-word self-contained blocks so AI can extract answers cleanly.

Google's position: there is no requirement to break content into tiny pieces for AI to understand it. Google's systems already handle multiple topics on a single page and surface the relevant part. Google engineers have specifically recommended against fragmenting content, and the guide states there is no ideal page length.

3. Rewriting content specifically for AI

The pitch: adopt a special "AI-friendly" writing style tuned for language models.

Google's position, verbatim: "You don't need to write in a specific way just for generative AI search."

4. Special AI schema

The pitch: add bespoke schema.org markup that AI engines reward.

Google's position: structured data is not required for generative AI search, and there is no special schema.org markup you need to add.

An important qualification, because this is the tactic most likely to be misread: Google is saying schema is not required for AI features. Structured data still earns rich results in regular Search and remains genuinely useful for that reason. The correct conclusion is "do not buy special AI schema as a GEO product," not "remove your schema."

The Two Tactics Google Warns Against

5. Manufactured brand mentions

The pitch: seed your brand across forums, listicles and community platforms so AI systems cite you.

Google's position: it explicitly warns against manufacturing inauthentic mentions to game generative AI results. Google's spam systems already filter much of what AI features depend on.

Why this matters commercially: genuine third-party presence does correlate strongly with AI visibility — that part of the GEO thesis holds. What Google is warning against is manufacturing it. Earned mentions and planted ones are different things, and the systems are built to tell them apart.

6. "Approved by Google" claims

Google's position, verbatim: some third-party services may make claims or imply that what they do is somehow "acceptable" or "approved" by Google Search — Google does not evaluate third-party services.

If a vendor implies Google sanctions their method, that claim is false by definition.

The Sentence That Applies to Every GEO Tool

The sharpest line in the June documentation is on the third-party tools page:

"Third-party tools don't have access to our internal ranking data. They can't guarantee performance. Any predictions are their own and like predictions generally, may not happen."

This is general language and it applies to every optimisation tool — established SEO platforms and newer AI-visibility products alike. No external tool sees Google's ranking signals, AI grounding logic, or the internal scoring that decides what appears in an AI answer. A tool reporting an "AI visibility score" is modelling, not measuring.

The practical implication for a Singapore business evaluating GEO proposals: any pitch promising a specific number — "we'll get you cited in X% more AI Overviews," "guaranteed AI Mode placement" — is making a claim Google says is structurally impossible for an outside tool to substantiate. The honest version of the pitch is probabilistic: do the fundamentals well and improve the odds.

What Google Says Actually Works

The tactics that survive the mythbusting are unglamorous, which is the point.

Unique, valuable, crawlable content. Google's stated position is that best practices for SEO continue to be relevant precisely because its generative AI features are rooted in core Search systems. The guide emphasises non-commodity content — material carrying a unique perspective, original research, and genuine first-hand experience that synthetic text cannot replace.

Technical hygiene. If pages are not correctly indexed due to technical problems, no content optimisation for AI features will help. Core Web Vitals, crawl budget and indexation errors in Search Console are the right starting point — not llms.txt or special markup.

Google Business Profile and Merchant Center. For local businesses and e-commerce, Google specifically names these as levers for AI visibility. This is not new advice, but it is a clear signal that Google treats its own products as primary data sources for AI answers in these segments. For a Singapore local service business, a well-maintained Business Profile is unavoidable.

The mechanics behind this, which the guide states openly: Google's AI features run on retrieval-augmented generation (grounding answers in real indexed content with links back to sources) and query fan-out (generating related concurrent queries to gather results before composing an answer). Both point to the same conclusion — indexed content quality drives AI visibility. If your page is genuinely useful, crawlable, and relevant to the fanned-out queries, it is a retrieval candidate. No amount of AI-specific markup substitutes for being a good answer.

What This Changes for a Singapore Business

If you are already being sold a GEO package, check it against this list. If the roadmap is built on llms.txt generation, content chunking or special AI schema, you are paying for work Google has publicly declared unnecessary for its AI features.

If you were planning to buy one, you now have a free public standard to evaluate it against. That standard did not exist before May 2026.

If you have already invested in the fundamentals — genuinely useful content, sound technical foundations, honest measurement — this guidance validates that work rather than undermining it.

What it does not mean is that AI visibility requires no attention. Google's guidance covers Google's surfaces. ChatGPT, Perplexity and Claude have different retrieval models, cite different sources, and show only limited overlap with each other. Measuring where you actually appear across platforms remains necessary work — it is simply not solved by the tactics on the debunked list.

Frequently Asked Questions

Q1: Did Google say GEO and AEO are pointless?

No — the opposite, with conditions. Google's May 2026 guide states that optimising for generative AI search is "still SEO" from its perspective, because its AI features are rooted in core Search ranking systems. Then on 5 June 2026, Google updated its "Do you need an SEO?" hiring guide to name AEO and GEO as legitimate services a credible SEO can provide — the first time that foundational document has done so. That is legitimisation, not dismissal. What Google did reject are specific tactics: six commonly sold techniques it says are either unnecessary or actively problematic.

Q2: Does llms.txt help with Google AI Overviews?

No. Google's guidance states verbatim: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search." Google may crawl an llms.txt file as it would any other page, but indexing a file does not mean it receives special treatment in Google Search. Some platforms charge monthly fees for "GEO file generation" — for Google's ecosystem specifically, that buys nothing. There is a more credible case for llms.txt with AI agents and documentation tooling, where adoption is growing, but not for Google's AI features.

Q3: Should I remove the schema markup from my website then?

No, and this is the tactic most likely to be misread. Google says structured data is not required for generative AI search and there is no special schema.org markup you need to add for AI features. That is different from saying schema is useless — structured data still earns rich results in regular Google Search and remains worth having for that reason. The correct conclusion is to stop buying "special AI schema" as a GEO product, not to strip existing markup from your site.

Q4: Can a GEO agency guarantee my business will appear in AI Overviews?

No, and Google has now said why in its own documentation: "Third-party tools don't have access to our internal ranking data. They can't guarantee performance. Any predictions are their own." No external tool sees Google's ranking signals, AI grounding logic, or internal scoring. A tool reporting an "AI visibility score" is modelling, not measuring. Any pitch promising a specific lift in AI citations is making a claim Google says is structurally impossible for an outside party to substantiate — treat guarantees as a disqualifier rather than a selling point.

Q5: Does Google's guidance apply to ChatGPT, Perplexity and Claude too?

No, and this is the scoping point most GEO pitches blur. Google's guidance covers Google's own AI features — AI Overviews and AI Mode — only. ChatGPT, Claude and Perplexity run on different indexing and retrieval models and may respond to factors Google dismisses. A vendor conflating "AI optimisation" across all platforms is over-generalising; the useful question to ask is which engine each specific recommendation is for. What holds across all of them is the unglamorous part: genuinely useful, crawlable content and sound technical foundations.

Mayson AI builds AI search visibility for Singapore businesses on fundamentals Google actually documents — not tactics it has publicly said are unnecessary. If you want an existing GEO proposal checked against the official guidance, book a consultation.

Quotations are from Google Search Central: "Optimizing your website for generative AI features on Google Search" (15 May 2026, revised 10 July 2026) and "Google Search's guidance on using third-party SEO tools, services, and advice" (5 June 2026). Verify current wording at developers.google.com/search.

The pitch: place a machine-readable manifest at your site root so AI systems can find and prioritise your content. Some platforms charge a monthly fee for "GEO file generation."

Google's position, verbatim: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search."

Google may crawl an llms.txt file as it would any page, but indexing a file does not mean it receives special treatment. For Google's AI features specifically, that monthly fee buys nothing.

The honest nuance: llms.txt is not dead everywhere — there is a more credible case for it with AI agents and documentation tooling, and adoption is growing in those contexts. But for the largest AI search surface in the world, Google has now said plainly that it does nothing.

The pitch: break long content into tiny 50-word self-contained blocks so AI can extract answers cleanly.

Google's position: there is no requirement to break content into tiny pieces for AI to understand it. Google's systems already handle multiple topics on a single page and surface the relevant part. Google engineers have specifically recommended against fragmenting content, and the guide states there is no ideal page length.

The pitch: adopt a special "AI-friendly" writing style tuned for language models.

Google's position, verbatim: "You don't need to write in a specific way just for generative AI search."

The pitch: add bespoke schema.org markup that AI engines reward.

Google's position: structured data is not required for generative AI search, and there is no special schema.org markup you need to add.

An important qualification, because this is the tactic most likely to be misread: Google is saying schema is not required for AI features. Structured data still earns rich results in regular Search and remains genuinely useful for that reason. The correct conclusion is "do not buy special AI schema as a GEO product," not "remove your schema."

Six: llms.txt and other AI-specific files, content chunking, rewriting content specifically for AI, special AI schema, manufactured brand mentions, and any vendor claim of being "approved by Google." On 15 May 2026, Google published its first official guidance on optimising for generative AI features — "Optimizing your website for generative AI features on Google Search" — and its most quoted section is titled, bluntly, "what you don't need to do." Google names these tactics directly, most with a verbatim "you don't need to" statement. Four are declared unnecessary; two are actively warned against. For any Singapore business currently being sold a GEO package, this document is the most useful procurement tool published all year, because it lets you check every line of a proposal against Google's own words. This article maps each tactic against the official position.

One scoping note before anything else, because most coverage blurs it: this guidance applies to Google's AI features only — AI Overviews and AI Mode. ChatGPT, Claude and Perplexity run on different indexing and retrieval models and may respond to things Google dismisses. Any vendor treating "AI optimisation" as one undifferentiated thing across all platforms is over-generalising.

The Framing: "Still SEO"

Google's central statement is worth quoting exactly, because it settles a debate that ran for two years:

"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."

The technical reason follows: Google's generative AI features are rooted in its core Search ranking and quality systems. The AI layer sits on top of the index, not beside it. The same index and quality signals that earn rankings are what earn citations in AI answers.

Much of the trade press read this as Google deflating the GEO category. There is a more accurate reading: on 5 June 2026, Google separately updated its "Do you need an SEO?" hiring guide to name AEO and GEO as legitimate services a credible SEO can provide — the first time that foundational document has done so. Google did not say GEO is fake. It placed GEO inside a discipline that already has standards and a public quality bar. That is legitimisation with guardrails, not dismissal.

The Four Tactics Google Says Are Unnecessary

1. llms.txt and AI-specific files

The pitch: place a machine-readable manifest at your site root so AI systems can find and prioritise your content. Some platforms charge a monthly fee for "GEO file generation."

Google's position, verbatim: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search."

Google may crawl an llms.txt file as it would any page, but indexing a file does not mean it receives special treatment. For Google's AI features specifically, that monthly fee buys nothing.

The honest nuance: llms.txt is not dead everywhere — there is a more credible case for it with AI agents and documentation tooling, and adoption is growing in those contexts. But for the largest AI search surface in the world, Google has now said plainly that it does nothing.

2. Content chunking

The pitch: break long content into tiny 50-word self-contained blocks so AI can extract answers cleanly.

Google's position: there is no requirement to break content into tiny pieces for AI to understand it. Google's systems already handle multiple topics on a single page and surface the relevant part. Google engineers have specifically recommended against fragmenting content, and the guide states there is no ideal page length.

3. Rewriting content specifically for AI

The pitch: adopt a special "AI-friendly" writing style tuned for language models.

Google's position, verbatim: "You don't need to write in a specific way just for generative AI search."

4. Special AI schema

The pitch: add bespoke schema.org markup that AI engines reward.

Google's position: structured data is not required for generative AI search, and there is no special schema.org markup you need to add.

An important qualification, because this is the tactic most likely to be misread: Google is saying schema is not required for AI features. Structured data still earns rich results in regular Search and remains genuinely useful for that reason. The correct conclusion is "do not buy special AI schema as a GEO product," not "remove your schema."

The Two Tactics Google Warns Against

5. Manufactured brand mentions

The pitch: seed your brand across forums, listicles and community platforms so AI systems cite you.

Google's position: it explicitly warns against manufacturing inauthentic mentions to game generative AI results. Google's spam systems already filter much of what AI features depend on.

Why this matters commercially: genuine third-party presence does correlate strongly with AI visibility — that part of the GEO thesis holds. What Google is warning against is manufacturing it. Earned mentions and planted ones are different things, and the systems are built to tell them apart.

6. "Approved by Google" claims

Google's position, verbatim: some third-party services may make claims or imply that what they do is somehow "acceptable" or "approved" by Google Search — Google does not evaluate third-party services.

If a vendor implies Google sanctions their method, that claim is false by definition.

The Sentence That Applies to Every GEO Tool

The sharpest line in the June documentation is on the third-party tools page:

"Third-party tools don't have access to our internal ranking data. They can't guarantee performance. Any predictions are their own and like predictions generally, may not happen."

This is general language and it applies to every optimisation tool — established SEO platforms and newer AI-visibility products alike. No external tool sees Google's ranking signals, AI grounding logic, or the internal scoring that decides what appears in an AI answer. A tool reporting an "AI visibility score" is modelling, not measuring.

The practical implication for a Singapore business evaluating GEO proposals: any pitch promising a specific number — "we'll get you cited in X% more AI Overviews," "guaranteed AI Mode placement" — is making a claim Google says is structurally impossible for an outside tool to substantiate. The honest version of the pitch is probabilistic: do the fundamentals well and improve the odds.

What Google Says Actually Works

The tactics that survive the mythbusting are unglamorous, which is the point.

Unique, valuable, crawlable content. Google's stated position is that best practices for SEO continue to be relevant precisely because its generative AI features are rooted in core Search systems. The guide emphasises non-commodity content — material carrying a unique perspective, original research, and genuine first-hand experience that synthetic text cannot replace.

Technical hygiene. If pages are not correctly indexed due to technical problems, no content optimisation for AI features will help. Core Web Vitals, crawl budget and indexation errors in Search Console are the right starting point — not llms.txt or special markup.

Google Business Profile and Merchant Center. For local businesses and e-commerce, Google specifically names these as levers for AI visibility. This is not new advice, but it is a clear signal that Google treats its own products as primary data sources for AI answers in these segments. For a Singapore local service business, a well-maintained Business Profile is unavoidable.

The mechanics behind this, which the guide states openly: Google's AI features run on retrieval-augmented generation (grounding answers in real indexed content with links back to sources) and query fan-out (generating related concurrent queries to gather results before composing an answer). Both point to the same conclusion — indexed content quality drives AI visibility. If your page is genuinely useful, crawlable, and relevant to the fanned-out queries, it is a retrieval candidate. No amount of AI-specific markup substitutes for being a good answer.

What This Changes for a Singapore Business

If you are already being sold a GEO package, check it against this list. If the roadmap is built on llms.txt generation, content chunking or special AI schema, you are paying for work Google has publicly declared unnecessary for its AI features.

If you were planning to buy one, you now have a free public standard to evaluate it against. That standard did not exist before May 2026.

If you have already invested in the fundamentals — genuinely useful content, sound technical foundations, honest measurement — this guidance validates that work rather than undermining it.

What it does not mean is that AI visibility requires no attention. Google's guidance covers Google's surfaces. ChatGPT, Perplexity and Claude have different retrieval models, cite different sources, and show only limited overlap with each other. Measuring where you actually appear across platforms remains necessary work — it is simply not solved by the tactics on the debunked list.

Frequently Asked Questions

Q1: Did Google say GEO and AEO are pointless?

No — the opposite, with conditions. Google's May 2026 guide states that optimising for generative AI search is "still SEO" from its perspective, because its AI features are rooted in core Search ranking systems. Then on 5 June 2026, Google updated its "Do you need an SEO?" hiring guide to name AEO and GEO as legitimate services a credible SEO can provide — the first time that foundational document has done so. That is legitimisation, not dismissal. What Google did reject are specific tactics: six commonly sold techniques it says are either unnecessary or actively problematic.

Q2: Does llms.txt help with Google AI Overviews?

No. Google's guidance states verbatim: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search." Google may crawl an llms.txt file as it would any other page, but indexing a file does not mean it receives special treatment in Google Search. Some platforms charge monthly fees for "GEO file generation" — for Google's ecosystem specifically, that buys nothing. There is a more credible case for llms.txt with AI agents and documentation tooling, where adoption is growing, but not for Google's AI features.

Q3: Should I remove the schema markup from my website then?

No, and this is the tactic most likely to be misread. Google says structured data is not required for generative AI search and there is no special schema.org markup you need to add for AI features. That is different from saying schema is useless — structured data still earns rich results in regular Google Search and remains worth having for that reason. The correct conclusion is to stop buying "special AI schema" as a GEO product, not to strip existing markup from your site.

Q4: Can a GEO agency guarantee my business will appear in AI Overviews?

No, and Google has now said why in its own documentation: "Third-party tools don't have access to our internal ranking data. They can't guarantee performance. Any predictions are their own." No external tool sees Google's ranking signals, AI grounding logic, or internal scoring. A tool reporting an "AI visibility score" is modelling, not measuring. Any pitch promising a specific lift in AI citations is making a claim Google says is structurally impossible for an outside party to substantiate — treat guarantees as a disqualifier rather than a selling point.

Q5: Does Google's guidance apply to ChatGPT, Perplexity and Claude too?

No, and this is the scoping point most GEO pitches blur. Google's guidance covers Google's own AI features — AI Overviews and AI Mode — only. ChatGPT, Claude and Perplexity run on different indexing and retrieval models and may respond to factors Google dismisses. A vendor conflating "AI optimisation" across all platforms is over-generalising; the useful question to ask is which engine each specific recommendation is for. What holds across all of them is the unglamorous part: genuinely useful, crawlable content and sound technical foundations.

Mayson AI builds AI search visibility for Singapore businesses on fundamentals Google actually documents — not tactics it has publicly said are unnecessary. If you want an existing GEO proposal checked against the official guidance, book a consultation.

Quotations are from Google Search Central: "Optimizing your website for generative AI features on Google Search" (15 May 2026, revised 10 July 2026) and "Google Search's guidance on using third-party SEO tools, services, and advice" (5 June 2026). Verify current wording at developers.google.com/search.

The pitch: seed your brand across forums, listicles and community platforms so AI systems cite you.

Google's position: it explicitly warns against manufacturing inauthentic mentions to game generative AI results. Google's spam systems already filter much of what AI features depend on.

Why this matters commercially: genuine third-party presence does correlate strongly with AI visibility — that part of the GEO thesis holds. What Google is warning against is manufacturing it. Earned mentions and planted ones are different things, and the systems are built to tell them apart.

Google's position, verbatim: some third-party services may make claims or imply that what they do is somehow "acceptable" or "approved" by Google Search — Google does not evaluate third-party services.

If a vendor implies Google sanctions their method, that claim is false by definition.

The sharpest line in the June documentation is on the third-party tools page:

"Third-party tools don't have access to our internal ranking data. They can't guarantee performance. Any predictions are their own and like predictions generally, may not happen."

This is general language and it applies to every optimisation tool — established SEO platforms and newer AI-visibility products alike. No external tool sees Google's ranking signals, AI grounding logic, or the internal scoring that decides what appears in an AI answer. A tool reporting an "AI visibility score" is modelling, not measuring.

The practical implication for a Singapore business evaluating GEO proposals: any pitch promising a specific number — "we'll get you cited in X% more AI Overviews," "guaranteed AI Mode placement" — is making a claim Google says is structurally impossible for an outside tool to substantiate. The honest version of the pitch is probabilistic: do the fundamentals well and improve the odds.

The tactics that survive the mythbusting are unglamorous, which is the point.

Unique, valuable, crawlable content. Google's stated position is that best practices for SEO continue to be relevant precisely because its generative AI features are rooted in core Search systems. The guide emphasises non-commodity content — material carrying a unique perspective, original research, and genuine first-hand experience that synthetic text cannot replace.

Technical hygiene. If pages are not correctly indexed due to technical problems, no content optimisation for AI features will help. Core Web Vitals, crawl budget and indexation errors in Search Console are the right starting point — not llms.txt or special markup.

Google Business Profile and Merchant Center. For local businesses and e-commerce, Google specifically names these as levers for AI visibility. This is not new advice, but it is a clear signal that Google treats its own products as primary data sources for AI answers in these segments. For a Singapore local service business, a well-maintained Business Profile is unavoidable.

The mechanics behind this, which the guide states openly: Google's AI features run on retrieval-augmented generation (grounding answers in real indexed content with links back to sources) and query fan-out (generating related concurrent queries to gather results before composing an answer). Both point to the same conclusion — indexed content quality drives AI visibility. If your page is genuinely useful, crawlable, and relevant to the fanned-out queries, it is a retrieval candidate. No amount of AI-specific markup substitutes for being a good answer.

If you are already being sold a GEO package, check it against this list. If the roadmap is built on llms.txt generation, content chunking or special AI schema, you are paying for work Google has publicly declared unnecessary for its AI features.

If you were planning to buy one, you now have a free public standard to evaluate it against. That standard did not exist before May 2026.

If you have already invested in the fundamentals — genuinely useful content, sound technical foundations, honest measurement — this guidance validates that work rather than undermining it.

What it does not mean is that AI visibility requires no attention. Google's guidance covers Google's surfaces. ChatGPT, Perplexity and Claude have different retrieval models, cite different sources, and show only limited overlap with each other. Measuring where you actually appear across platforms remains necessary work — it is simply not solved by the tactics on the debunked list.

Q1: Did Google say GEO and AEO are pointless?

No — the opposite, with conditions. Google's May 2026 guide states that optimising for generative AI search is "still SEO" from its perspective, because its AI features are rooted in core Search ranking systems. Then on 5 June 2026, Google updated its "Do you need an SEO?" hiring guide to name AEO and GEO as legitimate services a credible SEO can provide — the first time that foundational document has done so. That is legitimisation, not dismissal. What Google did reject are specific tactics: six commonly sold techniques it says are either unnecessary or actively problematic.

Q2: Does llms.txt help with Google AI Overviews?

No. Google's guidance states verbatim: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search." Google may crawl an llms.txt file as it would any other page, but indexing a file does not mean it receives special treatment in Google Search. Some platforms charge monthly fees for "GEO file generation" — for Google's ecosystem specifically, that buys nothing. There is a more credible case for llms.txt with AI agents and documentation tooling, where adoption is growing, but not for Google's AI features.

Q3: Should I remove the schema markup from my website then?

No, and this is the tactic most likely to be misread. Google says structured data is not required for generative AI search and there is no special schema.org markup you need to add for AI features. That is different from saying schema is useless — structured data still earns rich results in regular Google Search and remains worth having for that reason. The correct conclusion is to stop buying "special AI schema" as a GEO product, not to strip existing markup from your site.

Q4: Can a GEO agency guarantee my business will appear in AI Overviews?

No, and Google has now said why in its own documentation: "Third-party tools don't have access to our internal ranking data. They can't guarantee performance. Any predictions are their own." No external tool sees Google's ranking signals, AI grounding logic, or internal scoring. A tool reporting an "AI visibility score" is modelling, not measuring. Any pitch promising a specific lift in AI citations is making a claim Google says is structurally impossible for an outside party to substantiate — treat guarantees as a disqualifier rather than a selling point.

Q5: Does Google's guidance apply to ChatGPT, Perplexity and Claude too?

No, and this is the scoping point most GEO pitches blur. Google's guidance covers Google's own AI features — AI Overviews and AI Mode — only. ChatGPT, Claude and Perplexity run on different indexing and retrieval models and may respond to factors Google dismisses. A vendor conflating "AI optimisation" across all platforms is over-generalising; the useful question to ask is which engine each specific recommendation is for. What holds across all of them is the unglamorous part: genuinely useful, crawlable content and sound technical foundations.

Mayson AI builds AI search visibility for Singapore businesses on fundamentals Google actually documents — not tactics it has publicly said are unnecessary. If you want an existing GEO proposal checked against the official guidance, book a consultation.

Quotations are from Google Search Central: "Optimizing your website for generative AI features on Google Search" (15 May 2026, revised 10 July 2026) and "Google Search's guidance on using third-party SEO tools, services, and advice" (5 June 2026). Verify current wording at developers.google.com/search.