Ask ChatGPT the five questions your buyers ask before they shortlist a vendor, and see whether it names you, in what position, which competitors it recommends instead, and which sources it used. The score is free and needs no signup.
A ChatGPT visibility checker answers one question: when a buyer in your category asks ChatGPT for recommendations, does your brand come up? This checker does it the direct way. It reads your homepage, works out what you sell, writes five prompts a real buyer would type, sends each one to ChatGPT with web search switched on, and records what came back. You get a mention rate, an average position, the list of brands ChatGPT named instead of you, the sources it cited, and the exact sentence it wrote about you when it did name you.
Nothing here is a readiness estimate or a keyword lookup. The answers are live ChatGPT answers, generated at the moment you run the check, which is why they can differ from a check you ran last month and why we show the date on every report. If you want to know how crawlable and citable your site is, that is a different measurement and we have a separate tool for it further down this page. This one measures the outcome: are you in the answer or not.
A growing share of software and services buying starts with a question to an AI assistant rather than a search box. The buyer types "best CRM for a 20-person sales team" and gets a shortlist of five names with a reason for each. Most of them never open a second tab. If your brand is on that list, you are in the evaluation; if it is not, you are competing for a deal you were never invited to. That is a different problem from ranking, and it needs a different measurement.
The commercial effect shows up in two places. Inbound demo requests skew toward the brands ChatGPT names, because those brands are being shortlisted before anyone visits a website. And sales cycles lengthen for the brands it does not name, because prospects arrive with a competitor's framing already in their head. Knowing your mention rate turns an invisible loss into a number you can put in front of leadership, with the competitors and the sources attached.
Enter your domain in the checker above. If you want to steer the prompts, add a short description of what you sell, for example "payroll software for restaurants" or "AI SEO agency for B2B SaaS". Pick the country your buyers are in, then press Check ChatGPT. The checker reads your homepage, writes the five buyer prompts, asks ChatGPT each one and counts the results. Most checks finish in 20 to 30 seconds because five live answers are being generated at once.
The verdict, score, competitor list and sources are free and visible immediately. The per-prompt detail, meaning every brand ChatGPT named in each answer, the sentence it wrote about you and the links it cited, opens after you leave your work email. Every check also creates a permanent report page you can bookmark, share with your team or send to an agency, and re-run whenever you want a fresh set of answers.
Six things, each answering a question a marketing lead or founder actually asks. The first four are free on the page. The last two open with your email.
How many of the five buyer prompts produced an answer that named your brand. Five of five means ChatGPT treats you as a default option in your category. Zero of five means buyers who ask ChatGPT never hear about you.
Where you appear among the brands named in each answer. ChatGPT lists options in an order, and buyers read that order the way they read a shortlist. Being named fourth of six is a very different outcome from being named first.
Every brand ChatGPT recommended across the five prompts, ranked by how often it came up. This is your real competitive set in AI search, which is often different from the one on your battlecards.
The websites ChatGPT pulled from when web search was used: review platforms, comparison articles, forums, vendor pages. These are the places where recommendations get decided, and the places you need to be present.
The full ordered list of brands in each individual answer, with your brand highlighted. This shows which prompt types you win and which you lose, for example strong on "best for enterprise" and absent on "best for small teams".
The exact sentence ChatGPT wrote when it named you, plus the links it cited for that answer. If the description is outdated or wrong, you now know what to fix at the source.
All of this rolls up into a ChatGPT visibility score from 0 to 100. Seventy points come from mention rate and thirty from position, so a brand that is always named but always last scores lower than one that is named in most answers near the top.
In the Arobis AI Shortlist Study 2026, the most-recommended and least-recommended vendors inside the same software category were 94 points apart on recommendation frequency across five AI engines. Buyers using ChatGPT see one shortlist, and it is short.
See the State of AI Search Visibility dataChatGPT answers a "which vendor" question in one of two ways. With web search on, which is the default for most shopping and comparison questions today, it runs a search, reads a handful of pages and composes a shortlist from what those pages say. Without search, it answers from what it learned during training, which is older and favours brands that were already widely written about. Our checker runs prompts with search on because that is what a buyer gets, and it tells you on every report whether search was used.
In the search-on case, the shortlist is only as good as the sources. Review platforms, "best X for Y" articles, community threads and well-structured vendor pages get read first. If those sources describe you clearly and consistently, you get named. If they do not mention you, or describe you with the wrong category, you are absent no matter how strong your product is. That is why the sources tile on your report matters as much as the score: it shows which sites are shaping the answer for your category right now.
Three things move the outcome. Entity clarity: does every page that mentions you say the same thing about what you are and who you are for. Third-party presence: are you on the lists and review sites ChatGPT reads for your category. Consistency: the same prompt asked five times will produce five slightly different answers, so being named "sometimes" is normal and the useful measure is the rate, not a single answer. The checker asks five different prompts for exactly this reason.
Take a fictional example: a payroll product for restaurants, checked from the United States. ChatGPT names it in one of the five prompts, in fourth position, on the "best payroll software for small restaurants" question. On the other four prompts it names the same three large payroll brands every time, and cites two review platforms and one comparison article across all five answers. The score lands at 22, in the Invisible band.
The verdict is not "the product is bad". It is that the three sources ChatGPT reads for this category do not list the product, and the one that does describes it as "restaurant scheduling", not payroll. The fix is concrete: get listed on those two review platforms under the right category, pitch inclusion in the comparison article, and align the homepage description. That is the kind of specificity a generic visibility score cannot give you, and it is why the sources and the sentence ChatGPT wrote are part of every report.
Seventy or higher means ChatGPT recommends you in most buyer prompts and usually near the top. Here is how the four bands read and what each one usually means for pipeline.
Named in at most one prompt, or named late in the list. Buyers who start with ChatGPT are shortlisting competitors before they know you exist. Start with third-party presence: lists, reviews and comparisons.
Named in two or three prompts, often mid-list. You win some prompt types and lose others. The per-prompt detail shows which use cases ChatGPT associates with you and which it does not.
Named in four or five prompts, usually in the top three. The remaining gains are position and the description ChatGPT uses for you. Check the sentence it writes and fix outdated claims at the source.
Named first or second in nearly every prompt. Your job is to keep it: re-check monthly, watch which sources are cited, and make sure new competitors do not replace you in the lists ChatGPT reads.
Scores move between runs, and that is expected. ChatGPT samples its answers, web search results shift as pages are published, and a brand that is named in three prompts today can be named in four next week without anything changing on your side. Treat a single check as a snapshot and a series of monthly checks as the trend. A ten-point swing is noise; a band change that holds for two checks in a row is signal. The competitor list is the most stable part of the report, because the same three or four brands tend to dominate a category's sources for months, and it is the part worth acting on first.
Three tools with similar names answer three different questions. Use the one that matches the decision in front of you.
Arobis AI, free, above on this page
Answers
Does ChatGPT name my brand when buyers ask, and who does it name instead?
Reads
Five live ChatGPT answers to buyer prompts written for your category
Output
Mention rate, position, competitors, sources, the sentence about you
Time and price
About 25 seconds, free, no signup for the score
Use it when
You need to know whether AI search is sending buyers to you or to a competitor
Arobis AI, free AI visibility checker
Answers
Can AI engines crawl, understand and cite my website?
Reads
Your own site's signals: crawler access, schema, structure, llms.txt, technical health
Output
A readiness score per engine and a prioritised fix list
Time and price
About 25 seconds, free. Run the AI visibility checker
Use it when
You want to fix what blocks AI engines from reading you before working on mentions
SE Ranking, Ahrefs, Semrush and similar
Answers
How is my mention rate trending over weeks, across many prompts?
Reads
Large prompt sets run on a schedule against several engines
Output
Dashboards, share of voice, history, alerts
Time and price
Ongoing subscriptions, with limited free previews
Use it when
You already show up and want to monitor movement at scale
Most teams run this checker first to learn whether there is a problem, use the technical checker to remove anything that stops engines from reading them, and only then decide whether a paid tracker is worth it. If ChatGPT names competitors and not you, tracking that gap weekly does not close it. Changing what the sources say does, which is the work our AI visibility program for SaaS is built around.
Every check produces a public report page. Here are six well-known brands checked on September 7, 2026, with web search on, from the United States. Scores move between runs, so treat these as snapshots.
Notice what the low scores have in common. Webflow and Notion are household names, yet ChatGPT reached for Squarespace, Wix, ClickUp and Coda when a buyer asked for a recommendation, because the comparison pages and review sites it read for those prompts lead with the other brands. Fame is not the same as being on the shortlist. That gap is what the report shows you for your own category.
Work these in order. Each step feeds the next, and the report tells you where to start: a low mention rate points to steps 1 to 3, a low position points to steps 4 and 5, a wrong description points to step 2.
Your report lists the sites cited for your category. Review platforms, comparison articles and community threads decide who gets named. Claim your listings, collect reviews where buyers already are, and pitch inclusion in the "best of" pages that keep showing up. Our guide to AI visibility walks through each source type.
ChatGPT names brands it can categorise with confidence. Use one category phrase, one audience and one differentiator on your homepage, your About page, your review profiles and your LinkedIn page. When the sentence ChatGPT writes about you is wrong, this is almost always why.
"Best X for Y", "X vs Z" and "how to choose X" pages are exactly the prompts buyers type. When your own pages answer them honestly, with competitors named fairly, ChatGPT can cite you as a source, not only list you as an option. Structure each page as an answer: the question in the heading, the answer in the first paragraph.
Search-on answers only include sites the engine can fetch. Allow GPTBot and the other AI crawlers in robots.txt, keep pages fast and server-rendered, and add Organization and product schema. A free llms.txt generator gives crawlers a clean summary of your key pages in two minutes.
The unlocked report shows which prompts named you and which did not. If you are absent on "for small teams" or "on a budget", publish and get cited for that use case specifically. Position rises when more sources describe you as a fit for the exact question being asked.
Set up AI referral tracking with our guide to tracking AI traffic to your website, then re-run this check monthly. Sources change, competitors publish, and a brand that was named first can drop to third within a quarter without anyone noticing.
You can do a version of this by hand. Open ChatGPT in a fresh session with memory off, ask "what are the best [your category] for [your buyer]", and note every brand named and the order. Repeat with a comparison question, a "how do I choose" question and a "recommend one for [constraint]" request. Log the results in a table with one row per prompt and one column per brand. That table is your mention rate.
The reason a single chat misleads is variance. Ask the same question twice and the list shifts; ask it while logged in with chat history and your own past conversations colour the answer. The checker controls for this by asking five distinct prompts in a clean session, with the same instructions each time, and by reporting the rate rather than any one answer. It also captures citations, which you cannot see consistently by hand. Use the manual method to spot-check a specific prompt that matters to you, and the checker to get the baseline.
Each engine builds its shortlist differently. Perplexity searches on every answer and cites heavily, so source presence matters most there. Gemini and Google AI Overviews lean on Google's index and reward the same pages that rank well in Google. Claude searches when asked and otherwise answers from training, which favours established brands. A brand can be recommended by ChatGPT and absent from Perplexity, or the reverse, which is why we report engines one at a time instead of blending them into a single number.
This checker covers ChatGPT, the engine with the largest share of buyer questions. Our ChatGPT SaaS brand visibility study shows how differently the engines treat the same category, and the AI Visibility Audit runs your buyer prompts across all of them with the sources behind each answer.
You track rankings and pipeline, but nobody on the team can say whether ChatGPT recommends you. Run the check, screenshot the verdict, and you have the slide that starts the AI search conversation with your CEO.
No budget for enterprise tracking. Five minutes here tells you whether AI search is a problem worth solving this quarter, and exactly which competitors are collecting the buyers you are losing.
Check a prospect before the call, a client after the work, or a portfolio company during diligence. Every check has a permanent link, so the evidence travels with you.
Each check generates five buyer prompts for the category detected on your homepage, or the category you type in. The prompts follow a fixed mix: two "best of" questions, one comparison, one "how to choose" and one recommendation with a constraint. Each prompt is sent to ChatGPT through the OpenAI API with web search enabled and the same instructions every time; if search is unavailable, the report says "model knowledge only". Brands and positions are extracted from the answers, the score weights mention rate at seventy percent and position at thirty, and results for a domain are kept for seven days before a fresh run replaces them. Reports show the date they were generated. Prompts and answers are never edited by hand.
Yes. Ask ChatGPT for the best tool, agency or product in almost any category and it answers with a named shortlist, usually three to six brands with a line on each. With web search on, the list comes from the pages it reads at that moment; without search, it comes from training data. Either way, buyers act on it the way they once acted on the first page of Google.
Three causes cover most cases. The sources ChatGPT reads for your category do not mention you, so there is nothing to draw on. Your brand is described inconsistently across the web, so ChatGPT cannot confidently place you in the category. Or your site blocks AI crawlers, so search-on answers cannot fetch you. The report's sources tile shows which of the three applies: if the cited sites never mention you, start there.
Yes. Enter any public domain. The prompts are written for that domain's category, so checking your two closest competitors alongside your own domain gives you a side-by-side view of who ChatGPT recommends for the same buyer questions. Each check produces its own permanent report page.
It is the same source, done more carefully. The checker asks five different buyer prompts in a clean session with fixed instructions, captures the citations, extracts every brand and its position, and stores the result with a date. Asking by hand is fine for one prompt; it cannot give you a rate, and a logged-in chat with memory is coloured by your own history.
Those are trackers: they run large prompt sets on a schedule and sell dashboards, history and alerts, with limited free previews. This checker is a free, one-time diagnosis that shows the verdict, the competitors and the sources in under a minute, with no account. Use it to find out whether you have a problem; use a tracker later if you want to watch the trend at scale.
Both, in that order. Every prompt is sent with web search enabled, which is how ChatGPT handles most shopping and comparison questions for real users. If search is unavailable for a run, the checker falls back to the model's own knowledge and labels the report "model knowledge only", so you always know which kind of answer you are looking at.
Five prompts written for your category: two "best of" questions, one general and one with a concrete use case or company size, one comparison of the leading options, one "how do I choose" question, and one "recommend a ... for ..." request with a realistic constraint such as budget, region or team size. They never mention your brand, so ChatGPT is not led to it. The exact wording is shown on your report.
Seventy or higher. That means ChatGPT named you in four or five of the five prompts, usually in the top three. Between 40 and 69 you are named sometimes, typically mid-list, and below 40 you are effectively invisible to buyers who start with ChatGPT. The bands above explain what to work on at each level.
Be present, clearly described, on the sources ChatGPT reads for your category: review platforms, comparison articles, community discussions and your own answer-shaped pages. Use one consistent category phrase everywhere, publish honest comparison content, allow AI crawlers, and re-check monthly. The six steps above are the order that works.
Partly. With web search on, ChatGPT reads pages that a search engine surfaces, so pages that rank well are more likely to be read. But ranking your own site is not enough: the pages that shape a shortlist are usually third-party lists and reviews, not vendor homepages. A brand can rank first on Google for its category and still be absent from ChatGPT's answer if those lists leave it out.
No. llms.txt helps AI crawlers understand your site quickly and is worth publishing, but it does not put you on a shortlist. Recommendations come from what the web says about you. Treat llms.txt as hygiene, and third-party presence as the lever.
Monthly, and after any campaign that adds reviews, listings or comparison content. Results for a domain are kept for seven days; after that a new check runs fresh prompts. Because ChatGPT answers vary, compare mention rates over several checks rather than reacting to a single answer.
Any country. The picker lists every country and preselects the one your connection is in, so a visitor in France sees France by default and can change it with one click. The country is added to the prompts as the buyer's location, so ChatGPT recommends vendors relevant to that market. Most B2B software categories get similar shortlists across English-speaking markets; local services differ a lot.
Yes. Arobis Lens, our free Chrome extension, shows how often ChatGPT and other engines recommend the brand behind any site you are browsing. This page needs nothing installed; the extension is for checking as you go.
The score, competitors and sources are free without an email. Leaving your work email unlocks the per-prompt detail and lets us follow up once with the findings. Report pages are public by design, like the AI visibility reports, so do not enter a domain you would not want listed. We never sell or share your email.
Usually 20 to 30 seconds, because five live ChatGPT answers are generated at once. Checks are rate-limited per network to keep the tool free for everyone; if you hit the limit, wait an hour or use the permanent report link from your earlier check.
