Are competitors named instead of you in AI answers? How to check

By Ranjit Bhinge, founder of Blur Studio · Updated · 12 min read

Short answer: To check whether competitors are named instead of you in AI answers, write 20 questions your buyers ask, put each one to ChatGPT, Perplexity and Google's AI Overviews several times, and log who is named each time. Report your visibility as a percentage of runs, not as a "rank", because the same question gives different answers each time. Ploy, a software subscription where an AI agent builds and edits websites, can run a daily version of this check, and paid tools like Ahrefs and Semrush can too, but a spreadsheet is enough to start.

Ranjit Bhinge

Your buyers now ask AI tools for recommendations, and you cannot tell whether the answer names you, your competitor or nobody.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

You can check this in an afternoon with a spreadsheet. List the questions your buyers really ask, run each one on the AI tools they use, repeat each run several times, and write down who is named, who is linked, and whether the description is right. Then turn the log into a percentage: in how many runs were you named? That is your AI visibility. Any single run proves nothing, because these tools give different answers to the same question.

This page is for owners and marketers who want a repeatable check, not a one-off screenshot. It covers the method, why single runs mislead, what free and paid tracking costs, what to do if a competitor is winning, and where Ploy fits as one option of several.

This page may contain affiliate links from which Blur Studio earns commissions.

Why one run tells you nothing

SparkToro, a company that makes audience research software, published a study on 28 January 2026 (last modified 28 June 2026). About 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI Overviews, which made 2,961 runs in total. The surveys ran for about seven days each in November and December 2025, using default settings. A second phase used 142 human-written headphone prompts and produced 994 responses.

The findings that matter for you:

  • There was under a 1 in 100 chance that ChatGPT or Google's AI gave the same list of brands twice in 100 runs. Claude was slightly more likely to repeat itself.
  • An identical order was under 1 in 1,000.
  • Visibility as a percentage was steadier than position. In one example a cancer hospital was named in 69 of 71 ChatGPT answers (97 percent), but was the top mention only 25 times.
  • For a brand design agency prompt, visibility ran between 30 and 40 percent.

The author, Rand Fishkin, says tools that report a "ranking position in AI" are "full of baloney", and suggests running prompts about 60 to 100 times and averaging. He also says to ask tool vendors to show their math.

Read the caveats the study itself gives. It is not peer reviewed. The authors are not professional researchers. One co-author, Patrick O'Donnell, works at Gumshoe.ai, a company that sells AI tracking, and SparkToro disclosed that. The models date from late 2025. Treat it as a strong warning about single runs, not as a final measurement.

So the rule for the rest of this page: measure visibility percentage over many runs, never "AI rank".

What these tools actually do when they answer

You need a plain picture of how each tool finds sources, because it changes what you check.

  • ChatGPT search (OpenAI's chat assistant with web search) ranks results with factors OpenAI does not name. Answers may include citations, and a Sources button lists them. To be eligible, a site must let OpenAI's OAI-SearchBot crawl it. OpenAI says results and citations can be incomplete, outdated or incorrect.
  • Perplexity (an AI answer engine) searches the web in real time and puts numbered citations on each answer.
  • Claude (Anthropic's AI assistant) decides when to search the web and includes citations and source links when it does. Anthropic runs three crawlers. Claude-SearchBot improves search results, and Anthropic says blocking it may reduce your visibility in user search results.
  • Google AI Overviews and AI Mode use "query fan-out", which means running several related searches. Google says there are no additional requirements to appear and no special schema or new files needed. The page must be indexed and eligible to show a snippet.

One more fact stops you chasing the wrong thing. Ahrefs studied 15,000 long-tail prompts (published 11 August 2025) and found that on average 11.9 percent of URLs cited by AI tools also ranked in Google's top 10. Perplexity was higher at 28.6 percent. About 80 percent of citations did not rank anywhere for the original query. Ranking well in classic search is not the same as being cited.

Step 1: build your prompt list

Write 20 questions in the words a buyer would use, not the words you use. Mix these types:

  1. Category questions: "best roofing contractor for storm damage in Tulsa".
  2. Problem questions: "my basement floods after rain, who fixes it".
  3. Comparison questions: "[your brand] vs [competitor]".
  4. Brand questions: "is [your brand] any good" and "what does [your brand] do".
  5. Price and process questions: "how much does a roof replacement cost".

Get the raw wording from your sales calls, your inbox and your Search Console queries. If you invent every prompt yourself, your set may not match real buyer questions. That is the main weakness of any home-made list, so refresh it each quarter.

Step 2: run the prompts and log the results

Use a fresh session each time, logged out or in a private window, so your history does not shape the answer. Run each prompt with web search on, because that is how most buyers get current answers. If you want, run a second pass with search off to see what the model "knows" about you. Ahrefs, which sells SEO tools, recommends the same basics in its own guide, and says manual testing is a good way to set a benchmark but rarely scales.

A sampling plan you can copy:

SettingStarter planStronger plan
Prompts1020
EnginesChatGPT, Perplexity, Google AI OverviewsAdd Claude and Google AI Mode
Runs per prompt per engine510
CadenceMonthlyWeekly for the top 5 prompts
Runs in one full pass1501,000

The starter plan is 10 prompts times 3 engines times 5 runs, which is 150 runs. The stronger plan is 20 prompts times 5 engines times 10 runs, which is 1,000 runs for one pass, and it is too much by hand. Fishkin's 60 to 100 runs per prompt is a target for tools. By hand, five runs per prompt will show big gaps, but it will not give you a precise number. State the run count in any report.

Columns for your spreadsheet:

  1. Date and time
  2. Engine and whether web search was on
  3. Prompt and run number
  4. Were you named? (yes or no)
  5. Were you linked or cited? (yes or no, with the page)
  6. Which competitors were named, in order
  7. Was the description of you accurate? (yes, partly, no, and what was wrong)
  8. Which sources were cited

Then calculate: visibility percentage = runs where you were named, divided by total runs, for each engine and each prompt group. Do the same for each competitor. Mentions, citations and accuracy are three different things. A competitor named with a wrong price description is a different problem from one named correctly while you are absent.

A worked example with made-up numbers

A small accounting firm runs 10 prompts, 5 times each, on 3 engines. That is 150 runs. These numbers are invented to show the arithmetic.

EngineRunsYou namedCompetitor A namedYour visibility
ChatGPT5083116 percent
Perplexity50142728 percent
Google AI Overviews5052210 percent
All three150278018 percent

What the firm can say: "In 150 runs between 3 and 10 November, we were named in 18 percent and Competitor A in 53 percent." What it cannot say: "We rank fourth in ChatGPT." The second sentence invents a stable list that the evidence says does not exist.

Then the firm reads the 80 runs where Competitor A was named and checks what it has that the firm lacks: a clear page on small business tax filing, a price guide, reviews on a third-party site. That points to the next fix.

Free versus paid tracking

All prices below are list prices from each vendor's page in November 2026, with the billing basis shown. Check the vendor page before you buy, because plans change.

ToolWhat it isListed price and limits
Spreadsheet by handFree, your timeFine for 10 prompts a month
Ahrefs Brand RadarAn add-on to Ahrefs, an SEO tool makerCustom prompts listed from $50 a month. Checks equal prompts times platforms. Check the vendor page for tiers and extra checks
SemrushAn SEO and marketing suiteAI features sit inside its "SEO + AI Search" plans. The Starter plan listed $199 a month ($165.17 a month billed annually) for 50 prompts daily and one domain. Check the vendor page
Peec AIA tool that tracks brand visibility in AI answersStarter listed at EUR 85 a month, billed monthly, for 50 prompts and 3 models, as of November 2026. Check the vendor page, as the currency can differ by region
Otterly.aiAn AI search monitoring toolLite listed at $29 a month, excluding tax, for 20 prompts, as of November 2026. Check the vendor page for how credits work, as it can show euros or dollars by region
ScrunchAn AI visibility platformCore listed $250 a month on monthly billing, 15,000 AI responses a month. Check the vendor page
ProfoundAn enterprise AI visibility platformEnterprise with custom pricing only; earlier lower plans stopped being sold in September 2026

A rule for when a paid tool pays for itself: if you want weekly numbers for 20 prompts on 3 engines, that is at least 60 checks a week, about 240 a month, before repeat runs. A tool that quotes checks, prompts or responses lets you do that arithmetic against your own list. If a vendor will not tell you how often it runs each prompt and what it counts as a "mention", you cannot judge the number it gives you. No independent check of these tools against hand runs is cited by the vendors, so test one against your own log before trusting it.

Where Ploy fits, as one option of several

Ploy's AI search visibility docs describe a saved workflow, called a Ploybook, named "Track Your AI Visibility". It creates a project with a dashboard that asks your tracked questions daily and records where your brand and competitors appear. It uses AI credits. Ploy's docs list six engines: ChatGPT search, Gemini, Claude, Perplexity, Google AI Overviews and Google AI Mode. Ploy's feature page lists five and leaves out AI Mode, so check which set you get.

The same docs set honest limits. AI answers change often, so judge trends over weeks. No tool can guarantee that an AI assistant names you. Results can differ from what people see. A second Ploybook, the "Agent readiness audit", gives a score from 0 to 5 and fixes gaps it can. Ploy can also draft comparison, "best tool for X" and how-to pages. Drafts stay private until you publish, and a person should review them.

Plan facts as shown on Ploy's pricing page in November 2026: Free is $0 (500 credits a day, up to 2,000 in total), Starter is $50 a month (4,000 credits), Pro is $300 a month (24,000 credits) and Enterprise is custom. Credits do not roll over. Ploy does not publish the credit cost of a tracking run, and it does not publish caps on how many questions or engines you can track, so test on the Free plan and watch your credit use. See our Ploy pricing breakdown and Ploybooks guide.

This option makes sense if you already host your site on Ploy or plan to. If your site lives elsewhere, a spreadsheet or a dedicated tracker is the better start, because the tracking is only one part of what Ploy does.

Step 3: if a competitor is winning, what to fix

Nobody can make an AI tool name you. Google's guide to optimizing for generative AI features says that doing so is still SEO, and lists things that are not required: special schema markup, new machine-readable files, chunking your content, and rewriting only for AI. It also lists inauthentic mentions as something to ignore, so do not fake reviews or planted posts. Work in this order:

  1. Let the bots in. Check your robots.txt file (the file that tells crawlers what they may visit). If it blocks OAI-SearchBot, Claude-SearchBot or PerplexityBot, those search tools cannot read you. Blocking training bots is a separate choice.
  2. Make sure the page exists. If the question is "who does X in Tulsa" and you have no page about X in Tulsa, write one.
  3. Answer first. Put the direct answer in the first paragraph, then the detail. Our guide to getting mentioned by ChatGPT and Perplexity has a template, and our piece on what to automate in SEO and AEO retainers shows which of these jobs a person should keep.
  4. Fix thin and unclear pages. State prices, areas served, and who you are for. Fix wrong facts everywhere you control them.
  5. Look at the sources the AI cites. If answers lean on a review site or directory where you are missing or wrong, fix your listing there.
  6. Re-run monthly and write what changed in your log.

On llms.txt, a proposed file that lists your key pages for AI tools: SE Ranking studied nearly 300,000 domains in October 2026 and found no correlation with citation frequency. Ahrefs log data reported in June 2026 found 97 percent of valid files got no requests. It is cheap and harmless, but unproven. Do it last, if at all.

Also keep a sense of scale. Samples reported by SparkToro found small AI traffic: for 100 sites over 12 months, ChatGPT sent 0.62 percent of traffic (Marie Haynes), and across 95 sites organic search brought about 39 times the AI traffic (SEOWorks UK). These are small, third-party samples, so check your own analytics. Fishkin suggests treating AI visibility as a brand channel, not a traffic channel.

When this is not worth doing

  • Your buyers do not use AI tools to choose, for example when work comes only from referrals or contracts.
  • Your site is not indexed in Google yet. Fix that first. Our guide to a site not showing on Google helps.
  • You only want a rank number to show a boss. There is no stable one.
  • You cannot review what AI drafts. Do not publish unread AI-written pages.

For the bigger picture of how SEO, AEO (answer engine optimization) and GEO (generative engine optimization) differ, see SEO vs AEO vs GEO and what AEO is.

How I'd approach this

I would not buy a tracking tool first. I would run the free log once, with 10 prompts, 3 engines and 5 runs each, and see whether there is a problem worth paying to monitor.

First-week plan:

  1. Day 1: Write 10 buyer questions from real emails and calls.
  2. Day 2: Run them. Fill the log. This takes a couple of hours.
  3. Day 3: Calculate visibility percentage per engine and for each competitor. Read the answers where a competitor beat you.
  4. Day 4: Check robots.txt and your top three pages against the fix list.
  5. Day 5: Fix the one clearest gap and set a monthly re-run date.

Decision rule: if you are named in under 20 percent of runs and a competitor is named in over half, fix the content gaps before paying for anything. If you want it automated and your site is on Ploy, try the tracking Ploybook on the Free plan and compare its numbers to your hand log. If the two disagree widely, ask the vendor to show how it counts. If you need a custom content system or ongoing care of the pages AI tools read, Blur Studio builds that: see what Blur Studio builds.

Related guide: Are ChatGPT and Other AI Bots Reading Your Site? How to Check.

A closely related page covers the same ground from another angle: Selling AI Search Reports to Clients: Options for Agencies.

For the next step, see Make Your Website Readable for AI Agents: Where to Start.

Related guide: A Brand Presentation Deck Made Quickly: Options.

Questions people ask

How do I check if ChatGPT mentions my business?

Ask it the questions your buyers ask, with web search on, in a fresh session. Repeat each question several times and log whether your name appears. One run proves nothing because answers change from run to run.

Is there such a thing as an AI ranking position?

Not a stable one. SparkToro's study found under a 1 in 100 chance of getting the same list of brands twice in 100 runs, and an identical order under 1 in 1,000. Report visibility as the percentage of runs where you were named.

How many times should I run each prompt?

SparkToro's author suggests about 60 to 100 runs per prompt for a solid average, which suits tools. By hand, five runs per prompt will show large gaps. State the number of runs in every report.

Do I need a paid tool to track AI mentions?

No. A spreadsheet works for 10 to 20 prompts a month. Paid tools save time and repeat runs at scale. Check the vendor page for current prices and ask how each counts a mention.

Does ranking on Google mean AI tools will name me?

Not necessarily. Ahrefs found on average 11.9 percent of URLs cited by AI tools also ranked in Google's top 10 across 15,000 prompts, with Perplexity higher at 28.6 percent. See SEO vs AEO vs GEO.

Do I need special schema or an llms.txt file?

Google says there are no additional requirements and no special schema for AI Overviews and AI Mode. Studies from SE Ranking and Ahrefs found no measured effect from llms.txt. It is harmless but unproven.

Can Ploy guarantee that AI tools will name me?

No. Ploy's own docs say no tool can guarantee an AI assistant names you, and that results can differ from what people see. Judge trends over weeks. See our Ploybooks guide.

How often should I re-check?

Monthly is enough for most small businesses. Check weekly only for a handful of key questions or after a big site change, and compare the same prompts each time.

More in this topic: Websites: AI answers, AI bots and AI search reports (4 guides) →

Sources

Ready for the next step?

A 30 minute call. If you do not need me, I will tell you.

Ranjit Bhinge You talk to me, Ranjit. I reply within a few hours.

Not sure what you need? Tell me what is going on. I will point you the right way, even if it is not me.

Blur Studio · 30 minute call

Book a 30 minute call

You talk to me, Ranjit. If you do not need me, I will tell you.

  • We talk about your business and what is going on.
  • You leave with a clear next step, even if it is not me.
  • I reply to every message within a few hours.
Loading the calendar…