How to check if ChatGPT mentions your brand in 2026
To check if ChatGPT mentions your brand, ask buyer questions in an unpersonalized temporary chat, run each three times, log mentions, links and first place.
To check if ChatGPT mentions your brand in 2026, ask it the questions your buyers ask, not your brand name, in a temporary chat with personalization switched off. Run each question at least three times, because answers change between runs, and record whether you are named, linked and listed first. Then repeat the same questions in Perplexity, Gemini and Google's AI Overviews.
One check proves little. In a January 2026 study by SparkToro and Gumshoe.ai, the chance of ChatGPT or Google's AI giving the same list of brands in two answers was under 1 in 100. How often each brand appeared across many runs held much steadier, and that share is what this method measures.
The check fits in an afternoon. To improve what you find, read how to rank in ChatGPT and AI search.
What to ask: buyer questions, not your brand name
The questions to put to ChatGPT are the ones a buyer asks before they know your name: "best accountant for freelancers in Leeds", "how do I stop my basement flooding", "alternatives to [competitor]". Asking "What is [your brand]?" only tests whether ChatGPT knows you exist, not whether it recommends you to a stranger.
- Category and place: "best [service] in [city]" or "top [product type] for [use]".
- Problem first: "how do I fix [problem]", for buyers who don't yet know what to buy.
- Comparison: "[competitor] alternatives" or "[competitor] vs [competitor]".
- Constraints: "[service] for small businesses under [budget]".
- Advice: "what to look for in a [provider]", which shows the brands an AI holds up as the example.
Pick 10 to 15 questions for a first round, spread across the five kinds, and write each as a full sentence with the details that matter to that buyer.
Take the wording from real buyers, because people phrase the same need very differently. When SparkToro asked 142 people to write their own prompt for headphone recommendations, the prompts had a semantic similarity of 0.081, where 1 means identical. Sales calls, customer emails, the queries in your Google Search Console and "People also ask" boxes are good sources.
Branded questions ("What does [brand] charge?") have a different job: checking that the facts an AI repeats about you are right. Keep them on their own tab and out of your visibility numbers.
How to ask so memory doesn't skew the answer in 2026
ChatGPT personalizes answers when Memory is on, so your everyday account is the worst place to test: it may already know you own the business. The cleanest setup in 2026 is a temporary chat with personalization off, or ChatGPT while signed out, which OpenAI says can also search the web. Location still shapes both.
OpenAI's memory help page says ChatGPT can draw on past chats, saved memories and custom instructions, and that memory can shape the search queries it writes. As of October 2026, its temporary chat page adds a catch: a temporary chat can use your memories unless you choose otherwise. Select Temporary, then Unpersonalized, before the first message. That chat uses no memory or custom instructions and stays out of your history.
Perplexity's help center says memory and previous searches are always off in incognito mode. In Gemini, use a temporary chat: Google said at their August 2025 launch that these chats aren't used to personalize Gemini. For AI Overviews and AI Mode, use a signed-out private window, or the "Try without" link Google shows in the footer of personalized results.
Location is the part you can't fully switch off. ChatGPT estimates a rough location from your IP address and may add it to its searches, so "restaurants near me" becomes a search for restaurants in your city, according to OpenAI's web search help page. A VPN changes that guess. Test from your sales area, or name the city in the question.
| Tool | Cleanest way to ask | What still shapes the answer |
|---|---|---|
| ChatGPT | Temporary chat set to Unpersonalized, or signed out | Location from your IP address, your plan, the current model |
| Perplexity | Incognito mode | The mode you pick, such as Pro Search or Research |
| Gemini | Temporary chat | General location, or precise location if you allow it |
| Google AI Overviews and AI Mode | Signed-out private window, or "Try without" | The location shown in the footer, language, device |
Why does ChatGPT give different answers every time?
ChatGPT gives different answers to the same question because each answer is generated fresh. When it searches, OpenAI says ChatGPT rewrites your question into one or more targeted searches, and a different rewrite can bring back different pages. The SparkToro study measured how much this moves brand recommendations.
About 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI 2,961 times in November and December 2025. ChatGPT and Google's AI gave the same list of brands in two answers less than once in 100 tries, and the same list in the same order about once in 1,000. Yet Bose, Sony, Sennheiser and Apple appeared in 55 to 77% of headphone answers. The study wasn't peer reviewed, and Gumshoe.ai sells AI tracking tools.
Answers also shift when the model changes. GPT-6 reached paid ChatGPT plans on October 7, 2026 and free plans a day later, and within days answers looked and linked differently, Search Engine Journal reported. Date every check.
How many times to run each question
For numbers you can rely on, SparkToro's Rand Fishkin suggests running a prompt "usually at least 60-100X" and averaging. By hand, run each question three times per tool, each in a fresh temporary chat, and treat the result as a rough signal. Three of three suggests you are a regular in that answer, none suggests you rarely appear, and anything between is close to a coin toss. Ten questions in four tools already makes 120 answers, so keep the first list short.
Mentioned, cited or recommended first: what to record
Being mentioned, cited and recommended first are three different results, and an AI answer can give you one without the others. A mention is your name in the text, a citation is a link to your site as a source, and recommended first means you are the first or top pick. Record each separately.
| Result | What it looks like | What it tells you |
|---|---|---|
| Mentioned | Your brand's name in the answer | The AI counts you among the options |
| Cited | A link to your site, in the text or the Sources list | Your pages are read and used, even if your name isn't in the text |
| Recommended first | Named first, or called the best choice | You lead the answer, but this changes most between runs |
| Absent | Competitors named, you aren't | The gap to work on: note who was named and which sources were cited |
The three move separately. In SparkToro's data, City of Hope appeared in 69 of 71 ChatGPT answers to a hospital question, yet it was the top mention in only 25. Count how often you come first; don't read one position as a rank.
Many answers carry no inline link at all. In the same report, Peec AI's Tomek Rudzki found that after GPT-6, ChatGPT put a direct link to a website in 42% of answers, up from about 8 to 10%. So open ChatGPT's Sources button too, which lists cited sources.
An AI brand visibility check spreadsheet
An AI brand visibility check needs one spreadsheet row per answer, not per question, so repeated runs stay visible. Use these columns, then total the results per tool and per question.
| Column | What to write | Example |
|---|---|---|
| Date and place | When and where you asked | 28 Oct 2026, Leeds |
| Tool and setup | Tool, mode, signed in or out | ChatGPT, temporary, unpersonalized |
| Question and run | Exact wording, run 1, 2 or 3 | best accountant for freelancers in Leeds, run 2 |
| Mentioned | Yes or no | Yes |
| Position | 1st, 2nd, 3rd, later or none | 2nd |
| Linked | Did it link to your site? | No |
| Others named | Every competitor in the answer | Brand A, Brand B |
| Sources cited | The sites the answer linked | A directory, a review site |
| Facts wrong | Anything incorrect about you | Old price |
From those rows, four numbers per tool tell the story:
- Mention rate: answers that name you, divided by all answers. Named in 12 of 30 is 40%.
- First-place rate: answers that name you first, divided by all answers.
- Link rate: answers that link to your site, divided by all answers.
- Top rival: the competitor named most often, and its mention rate on the same questions.
How to read your AI visibility results
Read your AI visibility results as patterns, not single answers, and compare yourself with the competitors named on the same questions. No public benchmark says what a good mention rate is, and each pattern below points to its own fix.
- Never named or linked: check access first with our list of AI crawlers, then look at the sites the answers cite. Those pages would need to mention you.
- Named in ChatGPT but missing from AI Overviews: Google only links pages that are indexed and eligible for a snippet in Search, so the fix runs through Google Search. See how to appear in Google AI Overviews.
- Linked but not named: your page was a source, not the answer. Make sure it says plainly what you offer.
- Described wrongly: correct the fact on your own pages, then on the profiles the answers cite.
- One competitor everywhere: the reviews and comparison articles cited for them are your list of places to earn a mention.
- A big jump between months: look for a model change before you change anything.
Two free reports add your side. Google Search Console's generative AI report, open to all sites since August 31, 2026, shows impressions in AI Overviews and AI Mode by page, country and device. Bing Webmaster Tools' AI Performance report, launched February 10, 2026, counts citations in Copilot and Bing's AI summaries. Neither shows who was named instead. For the visits, see ChatGPT traffic in analytics.
When to automate an AI visibility check
Automate the check when the manual version stops being repeatable: more than about 20 questions, several cities or languages, or a need to look weekly. Software can run each question the many times reliable numbers need. Before choosing a tool, ask how it collects answers (SparkToro lists whether API answers match what people see in the apps as an open question), how many runs and locations it uses, and whether it reports a mention rate or a single "rank". Several tools specialize in this; our alternatives page covers many of them.
How QuQi helps
QuQi's AI Visibility teammate runs this check every week. It asks Google's AI Overview, ChatGPT, Perplexity and Gemini the questions your buyers ask, records who is named when you aren't, and turns each gap into a brief for the Copywriter, who researches the topic, writes the article in your business's voice and publishes it.
Every batch runs in a chat you can open and step into, and one switch per kind of work decides whether it goes out on its own or waits for your OK. No tool can promise a mention, QuQi included. See pricing.
Questions people ask
Does ChatGPT give the same answer to everyone?
No. Memory, past chats and custom instructions personalize answers, the location ChatGPT reads from your IP address changes local results, and plans can get new models on different days. Even with all of that equal, SparkToro's January 2026 study found ChatGPT returned the same list of brands in two answers less than once in 100 tries. Judge visibility by how often you appear across many runs.
Does ChatGPT temporary chat use memory in 2026?
Only if you let it. As of October 2026, OpenAI's help page says you choose Personalized or Unpersonalized before the first message, and can't change it later. Personalized can use existing memories, custom instructions and plugins; Unpersonalized uses none. Neither creates or updates memories, the chat stays out of your history unless you save it, and OpenAI may keep a copy for up to 30 days for safety.
How often should I check my AI brand visibility?
Monthly is enough for a manual check, with the same questions, setup and number of runs, so the months compare fairly. Run an extra check after a major model change, such as GPT-6 reaching ChatGPT on October 7 and 8, 2026, before acting on any drop. Weekly checks make sense once they are automated and no longer cost you an afternoon.
Why doesn't ChatGPT mention my business?
There is no single reason, but two checks come first. Make sure OAI-SearchBot, OpenAI's search crawler, can reach your site; OpenAI says a site must allow it, and let OpenAI's published IP addresses through, to be eligible for ChatGPT search. Then look at the pages ChatGPT cites for your buyers' questions. If those reviews, directories and articles never name you, ChatGPT has little reason to.
Do ChatGPT ads count as a brand mention?
No. Ads appear below ChatGPT's answer and are labelled, and OpenAI says they run on systems separate from the chat model and don't influence answers. OpenAI started testing them in the US on February 9, 2026, for Free and Go users, and temporary chats don't show them. Log only the names inside the answer itself.
What is a good AI mention rate?
No public benchmark exists, so compare yourself with the competitors named on the same questions. Rates differ widely by category: in Gumshoe's data, cited in the SparkToro study, top headphone brands often reached 90 to 100% visibility, while answers about brand design agencies peaked in the 30 to 40% range. A rising rate against the same rivals matters more than any single number.