How To Get Your Business Recommended By AI Search

In May 2026, Google did something unusual. It published official documentation telling website owners which popular AI search tactics to ignore.

The list was specific. You do not need llms.txt files. You do not need to break content into chunks for AI to understand it. You do not need to rewrite content in a special style for AI systems. You do not need special schema markup. You should not pursue inauthentic mentions across the web. Google’s guidance states plainly that its Search systems do not use llms.txt files, and that creating them will neither help nor harm your visibility.

An entire consulting category had been selling several of those tactics for over a year.

That tells you most of what you need to know about the state of this field. There is a genuine shift happening in how people find businesses, and there is a large and profitable industry selling certainty about it that nobody actually possesses. This article separates the two. Where something is documented by the platforms, it is stated as documented. Where it is inference from third-party testing, it is labelled as inference. Where nobody knows, that is said too.

One promise up front that you should treat as a warning sign whenever you hear it elsewhere:

nobody can guarantee that ChatGPT will recommend your business.

There is no ranking to buy, no position to occupy, and no submission process. Anyone selling that outcome is selling something they cannot deliver.

What you can do is improve the evidence these systems have to work with. That is a real strategy with real returns, and it is considerably less mysterious than the market suggests.

The Systems Are Not One Thing

“AI search” gets used as a single category. It describes at least four different mechanisms with different rules, and conflating them produces bad decisions.

Traditional Google Search

Ranked blue links. Still enormous, still the largest single source of discovery for most businesses, and still governed by conventional SEO.

Google AI Overviews and AI Mode

Generated summaries appearing within Google Search. Google’s documentation is explicit about how these work: they use retrieval- augmented generation , described as grounding, which relies on Google’s core Search ranking systems to retrieve relevant pages from the Search index, then generates a response from that retrieved material with clickable supporting links.

They also use query fan-out , where the model generates a set of concurrent related queries to gather more information. Google’s own example: a query about fixing a weedy lawn might fan out into queries about herbicides, chemical-free weed removal, and weed prevention.

Two consequences follow directly. First, eligibility runs through the ordinary Search index. Google states that to appear in these features a page must be indexed and eligible to appear in Search with a snippet. If you are not indexed, you are not eligible. Second, because of fan- out, your page can surface for questions adjacent to the one the user typed.

ChatGPT search

Architecturally different in an important way. ChatGPT does not search the web for every query. It answers from its training data unless a query triggers retrieval, and retrieval is more likely on commercial, comparative, and recency-sensitive questions than on general informational ones.

When retrieval does trigger, OpenAI documents three separate crawlers with independent controls:

OAI-SearchBot surfaces sites in ChatGPT search features GPTBot collects content that may be used to train foundation models ChatGPT-User fetches pages in response to specific user actions These are controlled separately in robots.txt. This is the single most consequential technical fact in this article, and it is widely misunderstood.

If you block OAI-SearchBot, you are opting out of appearing in ChatGPT’s search answers.

Blocking GPTBot only opts out of training use. Plenty of published guidance conflates these two, and some of it states, incorrectly, that blocking OAI-SearchBot affects only training.

Many sites have blocked the wrong bot, or blocked all AI crawlers through a security rule during an unrelated incident and never revisited it.

Checking this takes five minutes and is the highest-value action in this entire article.

The distinction that matters most

Across all of these, there is a difference between being mentioned and being cited . A mention comes from the model’s trained knowledge and typically carries no link. A citation comes from retrieval and usually does. They are produced by different mechanisms and are influenced by different things: mentions by broad presence across the web over time, citations by being retrievable, crawlable, and useful for a specific query right now.

SEO, AEO, GEO: What the Terminology Actually Means

You will encounter three acronyms, and the confusion around them is worth resolving because it drives a lot of wasted spending.

SEO is search engine optimisation, the established discipline.

AEO , answer engine optimisation, and GEO , generative engine optimisation, are both newer terms describing work aimed at visibility in AI- generated answers.

Neither AEO nor GEO has a settled, agreed definition. Different agencies use them to mean substantially different things, and some use them mainly as a reason to sell a new retainer for work that overlaps heavily with existing SEO.

Google’s position is on record and unambiguous: from Google Search’s perspective, optimising for generative AI search is optimising for the search experience, and therefore still SEO. Google explicitly directs readers considering third-party AEO or GEO services to its guidance on evaluating third-party SEO advice.

That is Google’s view of Google’s own systems, and it should not be over-extended. ChatGPT and Perplexity are not Google, and they weight things differently. What is fair to say is this: for Google’s AI features, the answer is settled and it is “this is SEO.” For other platforms, the fundamentals still dominate, but there are genuine differences at the margin, principally around crawler permissions and freshness.

Practical translation if an agency proposes a GEO programme that looks like content quality, technical accessibility, entity clarity, and earned coverage, that is legitimate work under a new label. If it proposes special files, AI-specific rewriting, or bulk mention-building, Google’s documentation says those do not work for Google, and there is no credible evidence they work elsewhere.

What Google Says You Can Ignore

Because this section corrects the most widely sold misinformation, it is worth stating precisely what Google’s documentation says rather than paraphrasing loosely.

llms.txt and similar files.

Google states you do not need to create machine-readable files, AI text files, markup, or Markdown to appear in Google Search including its generative AI features, because Google Search does not use them. Maintaining one for other services is fine and will neither help nor harm Google visibility.

Chunking content.

There is no requirement to break content into small pieces. Google states its systems understand multiple topics on a page and can surface the relevant part. There is no ideal page length.

Rewriting content specifically for AI.

Google states you do not need to write in a particular way for generative AI search, because these systems understand synonyms and general meaning. You do not need to capture every keyword variation.

Pursuing inauthentic mentions.

Google addresses this directly. Its generative features can reflect what is said about products and services across the web, but seeking inauthentic mentions is less helpful than it appears, because core ranking systems focus on quality while other systems block spam.

Over-focusing on structured data.

Structured data is not required for generative AI search and there is no special schema you need to add.

Google still recommends it for rich results as part of ordinary SEO, which is a meaningfully different claim from the one commonly made.

Third-party tools claiming internal metrics.

Google warns explicitly that no third-party tool has access to its internal ranking or AI systems, and to be wary of any tool promising ranking success or claiming to use internal Google metrics.

One more that deserves its own note, because it is actively risky rather than merely wasteful:

publishing large volumes of AI-generated pages is not a strategy.

Google states that doing this primarily to manipulate rankings or generative AI responses violates its scaled content abuse policy, and adds that a high quantity of pages does not make a website higher quality or more relevant. This is not a neutral waste of money. It carries downside risk.

What Actually Influences Whether You Appear

Not every claim about AI search carries the same weight. Sorted by how much evidence there actually is:

Confidence	What it is	Why it sits here
Established	Be indexable and crawlable	Documented by both Google and OpenAI
Established	Content quality and originality	Google says this matters more than anything else in its guide
Established	Business information systems	Business Profile, correct and current details
Reasonably supported	Authority and third-party coverage	Being cited and quoted elsewhere
Plausible	Structure that makes extraction easy	Clear headings, direct answers
Uncertain	Almost everything else	Most “AI SEO” advice being sold today

Here the evidence base thins, so the labelling matters.

Established: be indexable and crawlable

Documented by both Google and OpenAI. For Google’s AI features, a page must be indexed and eligible to appear with a snippet. For ChatGPT search, OAI-SearchBot must be permitted.

Practical checks

Confirm your robots.txt permits OAI-SearchBot if you want ChatGPT search visibility
Confirm no WAF or rate-limiting rule is returning errors to legitimate AI crawlers, which is a common and invisible cause of absence
Confirm important content is not locked behind JavaScript that crawlers cannot execute
Confirm your site is verified in Search Console so you can see what is happening

Established: content quality and originality

Google’s guidance is direct on this. It states that creating content people find unique, compelling, and useful will likely influence your presence in generative AI search more than anything else in its guide.

Google draws a specific distinction worth internalising.

Commodity content is based on common knowledge that could have come from anyone, and Google’s own example is something like “7 Tips for First-Time Homebuyers.”

Non-commodity content provides expert or experienced takes beyond common knowledge, and Google’s contrasting example is a piece explaining why the writer waived an inspection and what it cost them on the sewer line. That distinction is the most useful practical guidance in this entire field, and it is free. If a language model could generate your page from general knowledge without your business existing, that page adds nothing to the evidence pool. If your page contains something only you know, because you did the work, served the clients, or ran the numbers, it does.

Established: business information systems

For local and commercial queries, Google directs businesses to Google Business Profile and Merchant Center, noting these help products and services be visible in AI responses as well as ordinary results. This is documented, structured, and directly under your control, and it is neglected surprisingly often.

Consistency of core business details across the web is the boring foundation here. Name, address, phone, hours, and service area should match everywhere. Inconsistency creates ambiguity for any system trying to establish that a single entity exists.

Reasonably supported: authority and third-party coverage

Multiple independent analyses through 2025 and 2026 found that AI systems disproportionately cite sources with strong established authority signals, and that they lean toward third-party coverage over brand-owned content when both are available.

Treat the specific numbers in these studies with caution. They come mostly from SEO vendors, they use varying methodologies, they cannot access the systems they measure, and they sometimes contradict each other. What is reasonable to conclude is directional:

being written about credibly elsewhere appears to matter more for AI visibility than it did for classical SEO.

The mechanism is intuitive. A system synthesising an answer about the best accountant in a city has to decide what to trust. Independent corroboration is stronger evidence than self-description.

Note the important distinction Google draws, though. Earned coverage from genuine work is different from manufactured mentions, and Google says explicitly that pursuing inauthentic mentions does not help.

Plausible: structure that makes extraction easy

Third-party research consistently reports that content answering a question clearly and early gets cited more often than content that meanders. Google’s guidance points in a compatible direction, recommending content organised into paragraphs and sections with clear headings, while explicitly saying this is for readers rather than for AI.

This is worth doing because it is good writing practice regardless. It is not worth doing as a mechanical formula. The evidence for specific structural tricks is weak, and Google’s documentation contradicts several of them directly.

Uncertain: almost everything else

Specific citation-rate percentages, precise weighting of individual factors, and any claim about how a particular platform ranks sources are unverified. The platforms do not publish weighting formulas. OpenAI has not published one. Google states that no third-party tool has access to its systems.

If someone quotes you a precise number about how ChatGPT selects sources, ask where it came from and what the sample was. The answer is usually a vendor study measuring outputs from outside the system, which is legitimate research but cannot establish mechanism.

Two Firms, One Question

The mechanics become concrete with a comparison. These are illustrative rather than real companies.

Both are accounting firms in the same city. Someone asks an AI assistant to recommend an accountant for a small business with international contractors.

Firm A has a five-page website. The homepage says it provides high-quality accounting services with a personal touch. There is a services list, an about page naming no individuals, and a contact form. Everything is true and none of it is specific.

Firm B has published a guide to paying overseas contractors compliantly, with worked examples and actual figures. It has three case studies naming the problem, the approach, and the outcome. Its team page names each accountant with qualifications and specialisms. It publishes pricing bands and explains what drives them. It has a comparison page explaining when its service is the wrong fit and who to use instead. It has an active Business Profile with recent reviews. Two of its partners have been quoted in trade publications on contractor tax rules. Now consider what a retrieval system has to work with. For Firm A, there is essentially one usable fact: an accounting firm with that name exists in that city. Nothing connects it to international contractors, small businesses, or any demonstrated expertise. For Firm B, there is a page directly addressing the query, evidence the firm has done this work before, named humans with credentials, corroboration from third parties, and current business details. Firm B is more likely to be surfaced. Not because it optimised for AI, but because it produced more evidence that it can do the specific thing being asked about. This is the whole strategy, stated plainly. These systems answer questions using available evidence. Businesses that have generated more specific, credible, retrievable evidence about what they actually do get recommended more often. Firms that describe themselves in language interchangeable with every competitor do not, because there is nothing to retrieve. Note also that everything Firm B did has independent value. The guide helps prospects. The pricing page reduces unqualified enquiries. The case studies help close deals. None of it is wasted if AI search changes tomorrow, which is the correct test for any investment in this area.

A Practical 6 to 12 Month Plan

Months 1 to 2: Fix access and foundations

Audit robots.txt for every AI crawler you care about, particularly OAI-SearchBot. Check server logs and WAF rules for AI crawlers being blocked or rate-limited by accident. Verify Search Console and, where available, use the Generative AI performance report to see how content is performing in Google’s AI features. Complete and correct your Google Business Profile. Standardise business details everywhere they appear. Fix indexation problems on pages that matter.

This phase is unglamorous and frequently resolves the actual problem. A business absent from AI answers because a security rule blocks the crawler will not be helped by any amount of content.

Months 2 to 4: Build evidence only you can produce

Identify the ten questions your best customers ask before buying, in their words rather than in keyword form. Write genuinely useful answers to each, drawing on what you know from doing the work.

Prioritise formats that carry inherent specificity: case studies with real numbers, pricing explanations, comparisons including where you are the wrong choice, and analysis of your own data if you have any. Name your people, with credentials and specialisms.

Apply Google’s own test to every page: could this have been written by anyone with a general knowledge of the field? If yes, it is commodity content and it will not distinguish you.

Months 4 to 8: Earn external corroboration

Contribute expertise to industry publications. Respond to journalist requests in your field. Publish original data if you have any, since original data is the most citable asset a small business can produce. Speak where your industry gathers. Build genuine partnerships that produce genuine mentions.

Ask satisfied customers for reviews consistently and respond to all of them. Reviews are third-party evidence, publicly visible, and directly retrievable.

Do not buy mentions or run link schemes. Google addresses inauthentic mention-seeking explicitly, and the downside risk is real.

Months 8 to 12: Measure and maintain

Track referral traffic from AI platforms in analytics, which is possible because some platforms pass identifiable referrer parameters. Use Search Console’s generative AI reporting for Google. Test your own priority questions across platforms periodically and record what appears, accepting that these outputs vary between runs and are not a stable ranking.

Update your most important pages as facts change. Refresh because the content is out of date, not on a schedule for its own sake.

Most importantly, evaluate against business outcomes rather than visibility metrics. Enquiries, qualified leads, and revenue are the point.

Appearing in an AI answer that generates nothing is not a result.

What Nobody Can Tell You

Some honest limits.

Nobody can guarantee a recommendation.

There is no submission process, no paid placement in organic AI answers, and no ranking position.

Results vary between identical queries.

These systems are probabilistic. The same question can produce different sources on different runs, which makes “tracking your ranking” a category error.

The systems change without notice.

Retrieval behaviour, crawler policies, and interfaces have all shifted repeatedly and will continue to.

Measurement is genuinely immature.

Some referral traffic is identifiable, and Google now provides reporting for its own AI features.

Beyond that, tools estimate from outside the systems they measure. They can be directionally useful and cannot be treated as authoritative.

Nobody knows the long-term traffic picture.

Whether AI answers substantially reduce clicks to websites over time is contested and genuinely unresolved. Publishers report declines. Google states that clicks from pages with AI Overviews tend to be higher quality, with users spending more time on the destination site. Both claims can be partially true. Anyone presenting either side as settled is overreaching.

The Summary

The businesses that get recommended by AI systems tend not to have an AI strategy. They have a business that is genuinely distinctive and a website that documents it clearly, on a technically accessible site, corroborated by third parties who have no reason to flatter them.

Everything else on the checklist is either foundational hygiene, such as checking your crawler permissions, or is contested, such as the specific structural tactics that vendors sell with more confidence than the evidence supports.

The single highest-leverage question is the one Google essentially asks in its own documentation:

if you removed your company name from your website, would anything on it still identify your business as the right answer to a specific question?

If not, that is the work.

It is the same work that would win you referrals from humans, and it does not become obsolete when the technology shifts again.

Frequently Asked Questions

Can I pay to appear in ChatGPT or AI Overviews?

Not in the organic answers. There is no paid placement to buy for recommendations, and anyone offering it is misrepresenting what they can do.

Should I create an llms.txt file?

Google states it does not use them and that having one will neither help nor harm your Google visibility. If you want one for other services, it is harmless. It is not a strategy.

Does blocking AI crawlers hurt my business?

Blocking OAI-SearchBot removes you from ChatGPT search answers. Blocking GPTBot only affects training use. These are independent decisions, and many sites have blocked the wrong one by accident.

Do I need special schema markup for AI search?

Google says structured data is not required for its generative AI features and there is no special schema to add. It remains worthwhile for rich results as part of normal SEO.

How long until this shows results?

Technical fixes can affect eligibility within days. Content and authority work operates over months. Anyone promising rapid recommendations is not describing how these systems work.

Is traditional SEO dead?

No. Google’s AI features are built on its core Search ranking systems and require pages to be indexed and eligible for Search. The foundation is the same foundation.

Sources and Further Reading

Google Search Central, Optimizing your website for generative AI features on Google Search (updated July 2026). RAG and query fan-out, the AEO/GEO position, the mythbusting section, and the commodity versus non-commodity content distinction.

Google Search Central, AI features and your website . Eligibility requirements and preview controls.

Google Search Central, Guidance on third-party SEO tools and advice .

OpenAI, Overview of OpenAI Crawlers (developers.openai.com). Independent controls for OAI-SearchBot, GPTBot, and ChatGPT-User.

Google Search Console Help, Generative AI performance report .

Independent citation-behaviour analyses published by SEO research firms through 2025 and 2026. Directionally useful, methodologically varied, and not able to access the systems they measure. Treated in this article as inference rather than fact.

1. SEO title How to Get Your Business Recommended by AI Search (48 characters) 2. Meta description What actually influences whether ChatGPT, Gemini and AI search recommend your business, what Google says to ignore, and a realistic 6 to 12 month plan. (152 characters) 3. Suggested URL slug get-business-recommended-by-ai-search 4. Primary keyword get business recommended by ChatGPT 5. Secondary keywords

AI search optimization for business
generative engine optimization GEO
answer engine optimization AEO
how to appear in AI Overviews
ChatGPT search visibility
llms.txt do I need it
OAI-SearchBot robots.txt
AI search vs SEO

6. Suggested internal-link article ideas 1. AI Agent vs Chatbot: What Does Your Business Actually Need?

2. Why Most AI Automation Projects Fail and How to Avoid It 3. How to Audit Your Site for AI Crawler Access in One Afternoon 4. Case Studies That Actually Convert: A Practical Format 5. Original Data as a Marketing Asset for Small Businesses 7. Suggested CTA for The GosAI Most businesses missing from AI answers have a technical blocker, a content problem, or both, and the two need very different responses. If you would like your site checked against what the platforms actually document rather than what the market claims, get in touch with The GosAI.

Businesses that have generated more specific, credible, retrievable evidence about what they actually do get recommended more often. Firms that describe themselves in language interchangeable with every competitor do not, because there is nothing to retrieve.
The GosAI

If you would like an honest assessment of what your business would need to become the kind of source these systems can actually cite, get in touch with The GosAI.

How to Get Your Business Recommended by ChatGPT, Gemini and AI Search