AI visibility FAQ
Straight answers to the questions buyers, agencies and SEO leads actually ask about AI search: what it is, how to measure it, how to get cited, what to buy. 32 questions.
We publish an AI visibility platform, so we have an interest in this category. Where the honest answer is that a free check or a different kind of tool will do, we say so.
Jump to
Measuring AI visibility
What is AI visibility?
AI visibility is how often, and how favourably, an AI answer engine names your brand when a buyer asks a question in your category. It is measured by running a saved set of real buyer prompts across engines such as ChatGPT, Perplexity and Google AI Overviews, then recording whether you are mentioned, where you rank in the answer, which competitors appear beside you, and which pages get cited.
How do I check if ChatGPT mentions my brand?
Ask it. Write down ten questions a real buyer would ask before they know your name, run each one in a fresh conversation, and record whether you appear at all, who comes first, and which sources the answer links to. Do that once and you have a baseline. Do it on a schedule and you have a measurement.
How is AI visibility measured properly?
Three things have to be fixed before the number means anything: the prompt set, the engine list, and the record. Keep the same prompts, run the same engines, store the verbatim answer with its date, and compare like with like. Any tool that changes your prompt set between runs will show movement that is really just a different question.
What is share of voice in AI search?
Share of voice is your slice of the mentions across a fixed prompt set, relative to the competitors you track. If you are named in 12 of 40 answers and your three rivals are named in 30, 22 and 9, your share of voice is 12 of the 73 total mentions, about 16%. It is the single most useful number for a board or a client, because it does not depend on which prompts you happen to have picked.
How many prompts do I need to track?
Enough to cover the buying journey, not enough to drown in noise. For one brand, 20 to 50 real prompts spread across discovery, comparison and brand-verification questions is plenty to see movement. Our trial runs 10 so you can see how it works before paying.
Why do AI answers change from week to week?
Answers are assembled at request time from live search, model weights and rewriting, so small changes in what is indexed and how the model summarises it move the output. Volatility is the normal state. What matters is the direction of a fixed prompt over weeks, not one answer on one Tuesday.
Engines and how they differ
Which AI engines actually matter?
The ones your buyers use. For most B2B teams that is ChatGPT, Perplexity and Google AI Overviews, with Gemini and Copilot close behind. Tracking five engines badly is worse than tracking three well, because each engine sources its answer differently.
Does ranking in Google help me in ChatGPT?
It helps, but it is not the same ranking. ChatGPT's search-backed answers lean on pages that are crawlable, well structured and cited elsewhere, and Google AI Overviews favour pages that already rank. Strong classic SEO is a prerequisite rather than a guarantee.
How are Google AI Overviews different from ChatGPT answers?
AI Overviews are generated on top of a traditional Google result set, so they heavily reflect who already ranks and who is cited from those pages. ChatGPT answers are assembled from its own search and model knowledge, so a page that ranks position 12 in Google can still be the cited source there.
Why does the same question give different answers in different engines?
Each engine has a different index, a different set of trusted domains and a different summarisation style. Perplexity tends to cite densely, ChatGPT summarises and cites sparsely, and AI Overviews reuse Google's ranking. If you only measure one, you are measuring that engine's editorial taste, not your market.
Do the answers differ by country?
Yes, often sharply. The sources each engine trusts, the language of the answer and the competitors named all shift by market. Track the country your revenue comes from, not a global average.
Getting cited and fixing the gap
How do I get my brand cited by ChatGPT?
Make the page the model would want to quote: answer the question in the first two sentences, keep the claim verifiable, add the comparison table or the definition the answer needs, and make sure the page is crawlable by GPTBot and OAI-SearchBot. Then earn the third-party mentions that agents treat as corroboration. Being quotable and being corroborated are two different jobs, and both are required.
How long does it take to move an AI answer?
A technical fix can change what is crawled within days. A content change can show up in an answer in two to six weeks depending on how often that page is revisited. Authority work, the third-party sources an engine trusts, usually takes a quarter. Anyone promising next-week movement on a competitive prompt is guessing.
Does schema markup help AI visibility?
Schema does not make an AI cite you. Google's own guidance says structured data is not required for its generative AI features; what it buys you is machine-readable clarity and eligibility for rich results. Treat it as hygiene, not a lever.
Does an llms.txt file help?
For Google Search, no: Google states plainly that it ignores llms.txt and similar files. Some non-Google agents do read them, so serving one is cheap and harmless, but it is no substitute for crawlable HTML and pages worth citing.
Do backlinks still matter for AI search?
They matter differently. Classic links move rankings, which AI Overviews inherit. For ChatGPT and Perplexity, what matters more is being named on the pages the model already trusts: listicles, comparison posts, review sites and community threads. A brand mentioned in a Zapier roundup can outrank a brand with a thousand links and no mentions.
Should I rewrite a page or publish a new one?
Rewrite if the page already earns citations and merely answers the wrong question. Publish new if no page on your site addresses the prompt at all. The mistake to avoid is publishing a near-duplicate that splits the evidence between two URLs.
What is the single highest-leverage fix?
Find the prompt you lose where a competitor wins, read the source the engine cites in that answer, and become a better version of that source. It is unglamorous, it is specific, and it compounds because the same page often wins several prompts.
Can I measure whether a change worked?
Yes, if you fix the prompt before you ship. Record the verbatim answer, the ranking position and the cited sources, ship one change, then re-run the identical prompt after a defined interval. Keep the delta. Without the before-state, every later improvement is a story rather than evidence.
Tools, pricing and what to buy
What is an AI SEO tool?
AI SEO covers three jobs that are often sold as one: measuring how answer engines talk about you, optimising content so it is quotable and citable, and doing the technical and authority work that makes the page reachable. Most products do the first well, some do the second, few do the third.
How is this different from a normal SEO tool?
A normal SEO tool reports rankings on Google. An AI visibility tool reports whether you are named, recommended and cited inside a generated answer, plus which competitors appear beside you. They are complementary: classic rankings still influence AI Overviews, and AI answers still depend on crawlable pages.
What should I look for when buying one?
Four things: fixed saved prompts rather than ad-hoc checks, answer-level evidence with the verbatim text, competitor and share-of-voice tracking on the same prompt set, and citations so you can act. Ask how many prompts each tier runs, whether it tracks per country, and whether it stores the answers or just a score.
Is a free AI visibility checker enough?
A free checker is the right first step: it tells you whether you appear at all and whether your site is crawlable. Move to a paid monitor when you need the same prompts run on a schedule, competitor movement, and proof that a change moved the answer.
How much should AI visibility software cost?
Most categories settle between roughly $50 and $500 a month depending on prompt volume and engines, with enterprise contracts above that. Judge cost per tracked prompt and per engine, not the headline price, and check whether extra brands or clients are billed per seat.
How do I know if AI search is worth the budget?
Look at your own referral traffic first. If sessions from ChatGPT and Perplexity exist in your analytics and you cannot say which prompts produced them, the gap is measurement. If they are near zero, the work is earning mentions rather than buying a dashboard.
For agencies and multi-brand teams
Can agencies use this across client brands?
Yes. The Starter plan covers up to three brands with shared evidence and an approval trail, which is the shape a small agency needs: one prompt library per client, one report per client, and a record of what was changed and why.
Can I resell AI visibility reporting under my own brand?
Report on the client's own numbers, under your brand, is the normal arrangement. Whether you can resell the software itself as a white-label product depends on the plan you buy, so check the terms rather than assuming. What matters operationally is that each client has isolated projects and their own prompts.
How should an agency report AI visibility to a client?
Three slides: share of voice against named competitors on a fixed prompt set, the prompts lost and the sources that won them, and the work shipped with its verified delta. Clients renew on the movement of a number they have seen before, not on a dashboard they have never opened.
How do I prove ROI on AI search work?
Tie the prompt to a page, the page to an action, and re-run the prompt afterwards. When the same prompt later names the client and cites the page you shipped, that is the evidence. Everything else is correlation.
Technical basics
How do AI crawlers see my site?
Mostly as a plain HTML reader: no JavaScript execution, no cookies, no layout. If your content only exists after client-side rendering, or sits behind a login or a challenge page, the agent may see an empty shell. Check what a fetch without JavaScript returns.
What should robots.txt say about GPTBot and ClaudeBot?
Decide deliberately and write it plainly. If you want to be cited, allow the search and user-triggered agents such as OAI-SearchBot, ChatGPT-User, PerplexityBot and ClaudeBot, and consider what you want to do about pure training crawlers. Blanket-blocking everything is the fastest way to become invisible.
Do I need a separate site for AI search?
No. A second site splits the identity the model is trying to verify and is almost always a net loss. Fix machine-readability on the site that already has your brand, your domain and your links.