10 November 2026

AI search visibility: how to get your brand named and cited by ChatGPT, Google AI Mode, Gemini and Perplexity

What AI search visibility is, how answer engines pick the brands they name and the pages they cite, which crawlers to let in, and a 30-day plan with an honest way to measure it.

AI search visibility is how often AI answer engines such as ChatGPT, Google AI Mode, Gemini and Perplexity name your brand, and how often they cite your pages as a source, when people ask the questions your customers ask. To improve it, let the right crawlers in, publish pages that answer real questions with facts of your own, and be described accurately on the sites these engines already lean on. You measure it by asking the same questions again and again and counting, because one answer on one day proves very little.

This guide is for UK small businesses, online shops and the agencies that work for them. Vendor prices were checked on each vendor's own page on 10 Nov 2026 and are quoted in the currency the vendor publishes.

Key takeaways

  • AI search visibility has two parts: being named in the answer and being cited as a source. A brand can have one without the other.
  • Google says optimising for its AI features is "still SEO". No special file, markup or writing style is required for AI Overviews or AI Mode.
  • ChatGPT search and Perplexity each have their own search crawler. If your robots.txt blocks it, you are opting out of their answers.
  • Answers vary from run to run, so a single check is a snapshot. A percentage only means something with a sample size next to it.
  • Start with a fixed list of 10 to 20 customer questions, check them on a schedule, and log who is named, in what order, and which pages are cited.

What is AI search visibility?

AI search visibility is the share of AI-generated answers, for a set of questions, in which your brand appears. It is the answer-engine equivalent of ranking, except there is no fixed list of ten blue links. The engine writes a new answer each time, names some brands in the text, and attaches links to the pages it drew on.

That gives you three things to track:

  • Mentions. Is your brand named in the answer at all, and how early? An answer that lists five note-taking apps and puts yours fourth is not the same as one that leads with you.
  • Citations. Does the answer link to one of your pages as a source? You can be cited without being recommended, for example when your pricing page is used to describe a competitor's alternative.
  • Share of voice. Of all the brand mentions across your set of questions, what share is yours and what share belongs to each competitor?

People also call this work generative engine optimisation (GEO), answer engine optimisation (AEO) or LLM visibility. The labels overlap. Our piece on GEO vs SEO for a small business sets out what is different and what is ordinary search work with a new name.

How do AI answer engines decide which brands to name?

Most AI answer engines work in two steps: they run one or more web searches behind the scenes, then a language model writes an answer from the pages those searches returned, plus whatever the model already absorbed in training. Your brand gets into the answer through one of those two doors.

The search step is the one you can influence this month. Google describes its version as "query fan-out": AI Overviews and AI Mode may issue several related searches across subtopics and data sources, then build a response from the results (Google Search Central, AI features and your website). One customer question can therefore pull in pages that rank for narrower searches you never thought to target.

ChatGPT works in a similar way when it searches. OpenAI's help page says ChatGPT searches the web automatically when a question "might benefit from information on the web", shows inline citations in the reply and lists every source behind a Sources button (OpenAI Help Center, ChatGPT search). The same page says search queries may be shared with third-party search providers such as Bing, which is one reason Bing's index still matters to a UK business that thinks only about Google.

The training step is slower and harder to see. A model that read many consistent descriptions of your business before its cut-off date may name you without searching at all. You cannot edit that directly. What you can do is keep the public record about you accurate and consistent, so the next model learns the right thing.

Which AI engines matter for a UK business?

Five surfaces matter for most UK brands today: ChatGPT, Google AI Overviews, Google AI Mode, Gemini and Perplexity. Microsoft Copilot is a sixth where your customers use Windows and Bing at work. They do not behave alike, so treat each as its own channel.

Engine Where people meet it What controls whether your pages can appear First-party reporting for site owners
Google AI Overviews Above ordinary Google results, for some searches only Googlebot access, indexing and snippet eligibility Search Console
Google AI Mode A separate tab and google.com/ai; the UK is a supported country Same as AI Overviews Search Console
Gemini Google's assistant app and site Not documented separately for site owners; Google-Extended is Google's control for AI training and grounding in its other systems None published
ChatGPT App and site; searches when the question needs it OAI-SearchBot in robots.txt None published
Perplexity App and site; links to sites in its results PerplexityBot in robots.txt None published
Microsoft Copilot Windows, Edge, Bing Being indexed by Bing AI Performance in Bing Webmaster Tools (public preview)

Sources: Google Search Central, Google Search Help on AI Mode, OpenAI crawler documentation, Perplexity crawler documentation and the Bing Webmaster Blog, 10 Feb 2026, all read on 10 Nov 2026.

The two Google surfaces confuse people most. Google's own documentation says they "may use different models and techniques", so the set of links each shows will differ. Our comparison of AI Overviews vs AI Mode covers when each appears and what a brand should check on both.

Can AI crawlers actually read your site?

Check this first, because nothing else works if the answer is no. Each engine uses a named crawler for search, and a robots.txt rule or a firewall setting that blocks it removes you from that engine's answers.

Crawler Owner What it is for What blocking it does
Googlebot Google Google Search, including AI Overviews and AI Mode Removes pages from Search and its AI features
Google-Extended Google A control for AI training and grounding in some other Google systems Does not remove you from Google Search
OAI-SearchBot OpenAI Surfacing sites in ChatGPT search Site is not shown in ChatGPT search answers
GPTBot OpenAI Collecting content that may train OpenAI's models Signals no training use; separate from search
PerplexityBot Perplexity Surfacing and linking sites in Perplexity results Site may not appear in Perplexity results

Three details from the vendors' own pages. OpenAI says each robots.txt setting is independent, so you can allow OAI-SearchBot while disallowing GPTBot if you want to appear in answers without offering your content for training. OpenAI and Perplexity both say a robots.txt change can take around 24 hours to show up. Both recommend allowing requests from their published IP ranges, which matters if your host or CDN runs a bot-blocking rule you did not write.

For Google there is one more switch. Google's guide to optimising for generative AI says a site "must be included in Search generative AI features in Search Console" to be eligible to show in them, and points to a generative AI performance report for measuring it (Google Search Central, last updated 10 Jul 2026). Open Search Console and confirm the setting before you assume content is the problem.

A caution on llms.txt. It is a proposed file that lists your key pages for language models. Google's guide says Search does not need "new machine readable files, AI text files, markup, or Markdown" and ignores them. Adding one does no harm and other services may read it, but do not expect it to change what Google shows.

What makes a page worth citing?

A page gets cited when it holds a clear, specific, checkable statement that answers the question being asked. Engines are assembling an answer from parts, so the parts that travel well are short passages that make sense on their own.

Google's guide says optimising for generative AI search is "still SEO", that there is "no requirement to break your content into tiny pieces", and that you "don't need to write in a specific way just for generative AI search". It asks instead for content with a unique point of view from your own knowledge or experience, rather than recycled information that many other pages already carry.

The best public experiment on the other engines points the same way. A research team tested nine ways of rewriting web pages against a benchmark of 10,000 queries and measured how much of the generated answer each page won (Aggarwal et al., "GEO: Generative Engine Optimization", KDD 2024). On their main measure, adding quotations lifted a page's visibility by about 41%, adding statistics by about 31% and citing sources by about 27%. Keyword stuffing scored about 8% below the untouched page.

Read that study with care. Most of it ran on a test engine the authors built on an older OpenAI model, with a smaller check of 200 queries on Perplexity, and the authors found the effect varied by subject. It is evidence about direction, not a promise about your page. The direction is sensible enough to act on:

  • Answer first. Put the direct answer in the first two sentences under each heading, then explain.
  • Use your own numbers. Prices, lead times, sizes, test results, how many jobs you did last year. A figure that exists only on your site is a reason to cite your site.
  • Name things precisely. Say "our 14-day returns policy covers unworn items" instead of "we offer flexible returns". Models repeat specifics and skip vagueness.
  • Show sources and dates. Link the regulation, the standard or the dataset, and say when you checked.
  • Keep important content as text. Google lists this among its basics. A price that lives only inside an image or a script may never be read.
  • Match structured data to the page. Google says structured data should match the visible text. It does not ask for any special markup for AI features.

Why does the rest of the web matter as much as your own site?

AI answers about "the best" or "which should I choose" are usually built from third-party pages: comparison articles, directories, review sites, forums and trade press. If those pages describe your competitors and not you, the engine has nothing of yours to work with, however good your own site is.

So part of AI search visibility is ordinary, honest publicity. Make sure the places that already list businesses like yours list you correctly, with the same name, description, prices and location everywhere. Google's own checklist for AI features includes keeping Business Profile and Merchant Center information current. For a local or multi-branch business, reviews are part of that public record, and our guide on how to get more Google reviews sets out what Google's rules allow.

Do not try to shortcut this. Google's guide warns that "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem" and that its spam systems target manipulative tactics. Paid placements are legitimate when they are labelled as paid, and our explainer on rel="sponsored" vs nofollow covers how a paid link should be marked.

A practical way to find the pages that matter: look at what the engines cite today for your questions. If the same three comparison articles and one directory appear in answer after answer, those are the pages shaping what AI says in your category. Get accurately included there before you write new blog posts.

How do you measure AI search visibility?

You measure it by sampling. Pick a fixed set of questions, ask each engine the same questions on a schedule, record every answer, and count how often you are named and cited. Rank tracking does not carry over, because there is no stable position to track.

Three sources of data exist, and they answer different questions:

  • First-party reports. Google says sites that appear in AI features are counted in Search Console's Performance report under the Web search type, and its generative AI guide refers to a generative AI performance report. Bing Webmaster Tools has an AI Performance report in public preview that shows total citations, cited pages and sampled "grounding queries" across Copilot and Bing's AI summaries. These tell you about your own pages. They do not tell you which competitors were named.
  • Manual checks. Ask the questions yourself and write down the result. It costs nothing but time, and it is the right place to begin. Our walkthrough on how to track brand mentions in ChatGPT gives a spreadsheet layout and a weekly routine.
  • Tracking tools. Software that asks your questions every day on several engines and stores the answers. Our comparison of AI visibility tools lists published prices and what each plan includes.

Why does one answer prove so little?

The same question can produce a different list of brands an hour later, because the model's wording is partly random and the searches behind it change. Treat each answer as one draw from a bag.

A worked example, with invented figures. You ask a question 30 times over a month and your brand is named in 12 of the answers. Your visibility is 40%. But with only 30 answers, the 95% confidence interval for the true rate runs from roughly 25% to 58% (a Wilson score interval). If next month you score 47%, that is inside the same range, and you have not learned that anything improved. With 300 answers and the same 40%, the interval narrows to about 35% to 46%, and a move to 52% would be a real change.

So always write the number of answers beside a percentage. Compare periods only when each has a decent sample. Be suspicious of any report that shows a visibility score to one decimal place with no sample size.

Which questions should you track?

Track the questions a customer asks before they have heard of you. "Best accountancy software for a UK sole trader" is useful. "Is Acme Books any good" is not, because it names you and so tells you nothing about whether you come up unprompted.

Ten to twenty questions is enough for a small business. Take them from sales calls, support emails, the "People also ask" boxes on your main search terms, and the wording customers use in reviews. Write them as a person would say them, mix early research questions with buying questions, and keep the list fixed so months can be compared.

What should a small business do first? A 30-day plan

Spend the first month on access, a baseline and five pages, in that order: there is no point rewriting content an engine cannot fetch, and no way to judge a change without a baseline.

  1. Week 1: access. Read your robots.txt and your CDN or firewall bot settings. Confirm Googlebot, OAI-SearchBot and PerplexityBot are allowed. Check the generative AI setting in Search Console. Make sure your main pages are indexed in both Google and Bing.
  2. Week 1: questions. Write your 10 to 20 customer questions and three to five competitors.
  3. Week 2: baseline. Ask every question on ChatGPT, Google AI Mode and Gemini. Record the date, the engine, the brands named in order and the pages cited. Repeat at least three times across the week.
  4. Week 3: your pages. Pick the five pages that should answer those questions. Rewrite each so the answer comes first, your own facts and prices are in text, sources are linked and the page shows when it was last checked.
  5. Week 3: the public record. List the third-party pages the engines cited most. Where you are missing or described wrongly, ask for a correction or a listing, openly and with accurate details.
  6. Week 4: measure again. Repeat the baseline. Expect noise. Decide whether to keep checking by hand or hand the routine to a tool.

A shop already planning its year can fold this into the search section of its plan; AI answers are one more place where the same product facts need to be right.

What does not work, and what is risky?

Four things are sold as shortcuts and are not.

  • Writing for the robot. Google says there is no AI-specific writing style to adopt and warns that mass-producing pages to target every variation of a query can breach its scaled content abuse policy.
  • Keyword stuffing. It scored below the untouched page in the GEO study.
  • Buying mentions. Undisclosed paid mentions are the kind of manipulation search spam systems exist to catch, and they can mislead customers.
  • Guarantees. Nobody outside the engine controls its answers. Google's documentation says that meeting every requirement still does not guarantee a page is crawled, indexed or served. Treat any promise of a fixed place in ChatGPT as a warning sign.

One risk runs the other way: answers can be wrong about you. Google's help page says AI Overviews "can and will make mistakes". If an engine misstates your price or policy, find the page it cited, fix or ask for a fix at the source, and make the correct fact easy to find in plain text on your own site.

Doing this with CiteSieve

CiteSieve is Braivex's AI-search visibility tracker. It asks the questions you choose on real answer engines once a day and keeps every answer: the full text, the brands named in order, and the sources cited. You sign in at citesieve.braivex.com with a Braivex account.

What it does, as described on its own site on 10 Nov 2026:

  • Collects answers from ChatGPT, Google AI Mode and Gemini by default. Perplexity, Copilot and Google AI Overviews can be switched on.
  • Reports visibility as the share of answers that name you, with the number of answers and a 95% interval, and labels a week-on-week change as significant or within noise.
  • Shows share of voice for every brand an answer names, including ones you did not list, and ranks the sites the engines cited.
  • Runs a site audit that fetches your robots.txt, llms.txt, structured data, title and description. The audit uses no credits.
  • Lets you read any stored answer and export everything as CSV.

How it is priced: one credit is one answer from one engine. A question tracked daily on three engines uses about 90 credits a month. Brand plans are $95, $245 and $495 a month with 470, 1,200 and 2,400 credits; agency plans are $245, $495 and $795 a month with 1,200, 2,400 and 3,900 credits and client workspaces. The trial is 14 days with 70 credits and no card.

What it does not do. It cannot make an engine name you, and it does not write or change your pages. It does not show what real people typed into ChatGPT; it measures the questions you give it. Google shows an AI Overview for some searches only, so that engine returns nothing on many questions. In CiteSieve's own test on 10 Nov 2026, Copilot gave no answer for United Kingdom searches, so do not rely on Copilot figures for the UK. An engine that gives no answer is not charged.

Other Braivex tools for this job

  • Peakwyn is a paid, clearly labelled leaderboard where you can list a website; every rank is advertising, and our guide on where to list your startup or SaaS covers the free and paid places a new brand can be found.
  • ReelSieve turns an Airbnb listing into a short walkthrough video, which gives a holiday let something other than text for people to find.
  • HireSieve screens CVs against a rubric written from your advert, for the day the marketing work needs another pair of hands.
  • Reachwyn writes LinkedIn posts in the owner's own voice and publishes only what you approve.
  • Housieve scores UK sale, rental and commercial listings for agents and sourcers.
  • BriefSieve (VirtualPA) writes a morning brief from Shopify, Klaviyo and ad data; our ecommerce marketing plan shows where search fits in a small shop's year.
  • Loculens gathers Google reviews for every branch and drafts replies a person checks and posts; see our guide to multi-location reputation management.
  • VoiceSieve is Braivex's AI phone agent. It is in development and is not live.

Frequently asked questions

What is AI search visibility?

It is how often AI answer engines name your brand or cite your pages across a fixed set of customer questions, usually given as a percentage of answers with the sample size beside it.

How do I get my brand mentioned by ChatGPT?

Allow OAI-SearchBot in robots.txt, make sure your pages are indexed in Bing as well as Google, publish pages that answer customer questions with specific facts, and be listed accurately on the third-party pages ChatGPT already cites for your category. Nothing guarantees a mention.

Is GEO different from SEO?

For Google, no: its documentation says optimising for generative AI search is "still SEO". The differences are in measurement and in the extra crawlers used by ChatGPT and Perplexity.

Do I need an llms.txt file?

Not for Google, which says Search ignores AI text files. Other services may read one, so it is optional and harmless, but it is not a ranking or citation lever you can rely on.

How long does it take to appear in AI answers?

There is no published timetable. A robots.txt change can take about 24 hours to register with OpenAI and Perplexity. Content changes depend on recrawling, which Google says can take from several days to several months.

Can anyone guarantee a place in AI answers?

No. Google's documentation says that meeting every requirement still does not guarantee a page is crawled, indexed or served, and nobody outside an engine controls what it writes. Be wary of anyone selling a guaranteed mention.

How many questions should I track?

Ten to twenty unbranded customer questions is enough for a small business. Fewer questions asked more often give you a tighter result than many questions asked once.

If you want a first look before setting anything up, the free CiteSieve report runs a short one-off check on ChatGPT and Google AI Mode with a site audit and emails you the result in a few minutes. It is a snapshot, not a measurement, and it says so.