Skip to content

Search

AI search visibility: what a business can actually control in AI Overviews and AI answers

A new kind of pitch is doing the rounds: “AI search optimisation”, with an llms.txt file, a pile of FAQ markup and a promise that your business will be named when customers ask an AI assistant. Some of it is harmless. Some of it contradicts what the platforms themselves have published. This guide reads those publications, from Google, OpenAI, Perplexity and Microsoft, and separates what a business can control from what it cannot.

What decides whether Google’s AI Overviews use your page?

The same things that decide ordinary search. Google’s page on AI features and your website says: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” To be shown as a supporting link, “a page must be indexed and eligible to be shown in Google Search with a snippet”.

Google’s newer guide to optimising for generative AI search (last updated 10 July 2026) is blunter: “optimizing for generative AI search is optimizing for the search experience, and thus still SEO.” It names one factor above the rest: “Creating content that people find unique, compelling, and useful will likely influence your website’s presence in generative AI search in the long run more than any of the other suggestions in this guide.”

Read together, those lines make a plain point. If your service pages are thin, copied from a competitor or written around a keyword nobody types, no AI-specific trick will put them in an answer. If they say something specific and true that other pages do not, they have a chance, with no guarantee. Google’s first page says so too: “Indexing and serving isn’t guaranteed.”

Do you need llms.txt, special schema or “AI-ready” rewrites?

For Google Search, no. The optimisation guide is unusually direct about each one.

  • llms.txt and similar files: “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.” Keeping one is fine: it “will neither harm nor help your site’s visibility or rankings in Google Search, as Google Search ignores them.”
  • Structured data: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.” Google still recommends it for rich results, and the older AI features page asks that it match the visible text on the page.
  • Chunking pages into tiny sections and rewriting content “just for AI systems” are both listed as unnecessary.
  • Third-party tools: “No third-party tool has access to our internal ranking or AI systems.”

One more change matters to anyone paying for FAQ markup. Google’s Search Central changelog says the FAQ rich result “will no longer appear in Google Search starting May 7, 2026.” FAQ markup can still describe a page, but it no longer earns the expandable questions under a result.

We found no official statement from OpenAI, Perplexity or Microsoft that their search products read llms.txt. Their crawler pages explain robots.txt and IP addresses and say nothing about their search products reading llms.txt. The llms.txt proposal itself describes the file as a proposal.

What controls does Google give site owners?

Since 31 August 2026, a dedicated switch has been open to every site worldwide. Search Console’s Search generative AI control lets you include or exclude your site from AI Overviews, AI Mode and generative AI features in Discover. Including is “the default control for all properties.” Google says a change “generally takes a few days” to apply, and some content takes longer because of caching.

ControlWhat it doesWhat it does not do
Search generative AI control (Search Console)Keeps your links and content out of AI Overviews, AI Mode and Discover AI features, or lets them inNot a ranking signal for the rest of Search; does not stop AI training
nosnippet, data-nosnippet, max-snippetLimit what text Google may show from a page, which also limits what AI features can showDo not remove the page from Search
noindexRemoves the page from Google Search entirely, AI features includedCannot keep you in ordinary search; use the Search Console control if you only want out of AI features
Google-Extended (robots.txt token)Limits use of your content for training Gemini models, including those behind Search’s AI features, and for grounding in Gemini Apps and Vertex AIGoogle: it “does not impact a site’s inclusion in Google Search nor is it used as a ranking signal”

The last row trips people up. Blocking Google-Extended does not take you out of AI Overviews; Google’s crawler documentation says it “does not impact a site’s inclusion in Google Search”, and Google’s AI features page describes it as a control for “AI training and grounding in some of Google’s other systems”. The new Search Console control works the other way round: Google says it “doesn’t affect AI training”. For training, including the models behind Search’s AI features, Google points to Google-Extended instead.

Before excluding anything, read the warning on Google’s page for the control: “Content from other sites will still be available in those features, and it may appear similar to yours.” For most small businesses that want customers to find them, excluding is the wrong move. It suits publishers who sell their content.

How do you see AI traffic in Search Console?

Also since 31 August 2026, Search Console has a generative AI performance report. Google says “Impressions are how many times links to your site were shown to a user in a generative AI feature on Google Search”, and the report breaks them down by page, country, device and date. Google’s help page describes impressions; we did not find clicks documented there.

Clicks from AI features still sit inside the ordinary Performance report, in the “Web” search type, as Google’s AI features page explains. So the useful habit is simple: once a month, check which of your pages get AI impressions, and whether those same pages produce enquiries. If you have not yet set up enquiry tracking, our guide to lead tracking with UTM tags, GA4 and a CRM covers that.

What about OpenAI, Perplexity and Microsoft Copilot?

Each publishes its own crawler rules. The pattern is the same: one crawler decides whether you appear in search answers, another decides whether your content trains models, and fetches a user asks for directly may ignore robots.txt.

PlatformCrawler or settingWhat it controls
OpenAIOAI-SearchBotWhether your site can appear in search answers inside OpenAI’s assistant; OpenAI says opted-out sites “can still appear as navigational links”
OpenAIGPTBotWhether your content may be used to train OpenAI’s foundation models
OpenAIThe user-initiated fetcher (its user-agent name ends in -User)Pages fetched because a user asked; OpenAI says “robots.txt rules may not apply”
PerplexityPerplexityBotWhether your site is surfaced and linked in Perplexity search; Perplexity says it is not used for training
PerplexityPerplexity-UserFetches a user requests; it “generally ignores robots.txt rules”
Microsoft Bing and Copilotnoarchive and nocache meta tags; data-nosnippetHow much of a page Bing’s AI answers may use, per Microsoft’s 2023 Bing Chat rules; noarchive also opts content out of Microsoft’s model training (see below)

OpenAI’s crawler page says “Each setting is independent of the others”: you can allow OAI-SearchBot and disallow GPTBot. It recommends allowing OAI-SearchBot and OpenAI’s published IP ranges if you want to appear in its search results, and notes that a robots.txt change can take “~24 hours” to apply. Perplexity’s crawler page gives the same advice for PerplexityBot and warns that Cloudflare’s Bot Fight Mode can still challenge its bots even when you allow them.

That warning is the practical one for small sites, because a firewall or hosting plan may block bots it does not recognise. If you want to be found in these answers, ask whoever runs your hosting whether OAI-SearchBot, PerplexityBot and Bingbot are blocked.

Bing and Copilot

Microsoft set out its rules in a 2023 Bing Webmaster post, written when Copilot was still called Bing Chat. “Content tagged NOARCHIVE will not be included in Bing Chat answers”; with NOCACHE, “We will only display URL/Snippet/Title in the answer”. Pages with either tag “will still appear in our search results”, and the same post says the tags also govern use for training Microsoft’s models. Microsoft has not restated these rules for Copilot on a page we could read, so treat them as the best available guide, not a guarantee. In 2025 Bing added support for data-nosnippet, which keeps a marked section “excluded from snippets and AI summaries” while the page stays indexed.

Bing Webmaster Tools also has an AI Performance report, in public preview since 10 February 2026, which shows how often your pages are cited in Copilot and Bing’s AI summaries. Of the reports open to any site owner that we found, it is the only one that counts citations by page; Google’s report counts link impressions instead. We found no equivalent report from OpenAI or Perplexity open to every site. Two free signals come close: your server logs, read by crawler name, show which pages their crawlers fetch, and the AI Assistant channel in GA4’s default channel group shows visits that arrive from AI assistants such as Gemini and Copilot. GA4 counts clicks from Google’s AI Overviews and AI Mode as Organic Search instead.

Is AI search sending fewer visitors to websites?

A study with a published method comes from the Pew Research Center, published on 22 July 2025. It tracked the browsing of 900 adults in the United States during March 2025. When a Google results page showed an AI summary, people clicked a traditional result in 8 per cent of visits, against 15 per cent when there was no summary. They clicked a link inside the summary in just 1 per cent of visits.

Google reports a different picture for overall traffic. Its head of Search wrote on 6 August 2025 that “total organic click volume from Google Search to websites has been relatively stable year-over-year.” Pew’s sample is American and from early 2025; no comparable study of Indian users has been published that we could find.

The fair reading for a small business: questions with a quick factual answer (“what does a modular kitchen cost per foot”) are the ones most likely to be answered on the results page. Questions that end in a decision (“which kitchen company in Baner can install before Diwali”) still need your page, your reviews and your phone number. Put your effort into pages for the second kind.

How widely are AI answers used in India?

Google launched AI Overviews in India on 15 August 2024, in English and Hindi. AI Mode arrived on 24 June 2025 as an experiment in Labs, in English. Hindi followed, and on 8 October 2025 Google added Bengali, Kannada, Malayalam, Marathi, Tamil, Telugu and Urdu. In July 2025 Sundar Pichai said AI Mode already had “over 100 million monthly active users in the U.S. and India.” A year later, in his July 2026 remarks, he said AI Mode had “surpassed 1 billion monthly active users” worldwide and that Google had “brought together AI Overviews and AI Mode into one seamless Search experience”. These are Google’s own counts; we found no independent figure for India.

Marathi matters for anyone selling in Pune. If your customers search in Marathi or Hindi and your site is English only, our view is that the answer they see will more likely draw on somebody else’s page.

What should a business do this month?

  1. Confirm Google can crawl and index your important pages: run them through Search Console’s URL inspection.
  2. Check that your hosting or firewall does not block Googlebot, Bingbot, OAI-SearchBot or PerplexityBot, unless you have decided to block them. A robots.txt that welcomes search crawlers but refuses OpenAI’s training crawler can be as short as two lines: the first reads User-agent: GPTBot and the second reads Disallow: / (everything else stays allowed by default).
  3. Leave Search Console’s generative AI control on Include unless you have a clear reason, such as content you sell.
  4. Keep your Google Business Profile and, if you sell products, Merchant Center current. Google says both “can help your products and services to be visible in both AI responses and other Google Search results.”
  5. Rewrite your three most important service pages so each answers the questions a buyer asks before calling: what is included, for whom, how long it takes, where you work. Specific beats long.
  6. Add local-language pages where your customers search in Marathi or Hindi.
  7. Keep structured data accurate and matching the page. Do not add markup to chase AI answers.
  8. Set up Bing Webmaster Tools and check its AI Performance report alongside Google’s generative AI report each month, and look at GA4’s AI Assistant channel.
  9. Treat any offer of guaranteed placement in AI answers as a red flag. Google says no outside tool can see its systems.

Most of this is ordinary search work done well, which is the point Google keeps making. Our Google Business Profile guide covers the local half in detail.

QUESTIONS

Common questions

How do I get my business into Google’s AI Overviews?

Google says there are no additional requirements or special optimisations. A page must be indexed and eligible to show with a snippet. Beyond that, Google says unique, useful content matters most. No one can guarantee placement.

Does llms.txt help with Google’s AI answers?

No. Google says Google Search ignores llms.txt files, and that having one neither harms nor helps your visibility. We found no official statement from OpenAI, Perplexity or Microsoft that their search products use it.

Can I stop Google using my site in AI Overviews?

Yes. Since 31 August 2026, Search Console’s Search generative AI control lets you exclude your site from AI Overviews, AI Mode and Discover’s AI features. Blocking Google-Extended does not do this. It limits training of Google’s models, including those behind Search’s AI features, and grounding in Gemini Apps, but it does not affect whether your pages appear in Search.

How do I appear in OpenAI’s or Perplexity’s search results?

Both recommend allowing their search crawlers, OAI-SearchBot and PerplexityBot, in robots.txt and allowing their published IP ranges through any firewall. Allowing the search crawler is separate from allowing training crawlers such as GPTBot.

Is FAQ schema still worth adding?

Google stopped showing FAQ rich results from 7 May 2026, and says no special markup is needed for its AI features. Structured data that matches the page is still useful for other rich results.

REFERENCES

Sources and further reading

Each link was opened and checked on .

← All field notes