All articles
AI searchGEOAI visibility

How to Check Whether ChatGPT, Gemini, and Perplexity Recommend Your Business

Run a free AI search visibility check on ChatGPT, Gemini, and Perplexity with a 30-minute monthly protocol, a log template, and tools worth paying for.

On this page

Somewhere out there, a potential customer just asked ChatGPT which business like yours to hire. Whether your name showed up in the answer is something you can check yourself, today, without spending a dollar.

The AI search visibility check I’d run looks like this: build a fixed set of 10–20 buyer-style prompts, run each one across ChatGPT, Gemini, and Perplexity in fresh chats, log which brands get mentioned and which sources get cited, then repeat the whole thing monthly. A spreadsheet handles it fine. Monitoring tools just automate the same job once you outgrow the manual version.

Set aside 30 minutes each month to repeat the check. Save the answers and citations so you can see whether a change persists across runs.

Before you run anything: the answers change every time

One run of one prompt tells you almost nothing, because these models are built to vary their answers. Plan for repetition before you plan for anything else.

Don’t take my word for it. In January 2026, Rand Fishkin at SparkToro published research with Patrick O’Donnell of Gumshoe that put hard numbers on this. Around 600 volunteers ran the same 12 prompts through ChatGPT, Claude, and Google’s AI Overviews a combined 2,961 times.

The chance that ChatGPT or Google’s AI would give you the same list of brand recommendations twice was less than 1 in 100. The same list in the same order? Less than 1 in 1,000.

Even the length of the list moved around. Sometimes the model gave two or three recommendations, sometimes ten or more, for the identical prompt.

Why does this happen? These systems are probability engines, not databases. They sample from distributions, they fetch live web results at answer time (and fetch different ones each run), and providers quietly swap model versions without telling you, the variability isn’t a bug someone forgot to fix, it’s the design.

That forces two rules onto your method:

  • Fix your prompts. Write them once, save them verbatim, and never edit them mid-year. Rewording a prompt changes the answer, which poisons the comparison.
  • Sample repeatedly. One answer is an anecdote. Monthly repetition gives you a trend, and trends survive the noise that single answers drown in.

One more thing the SparkToro study settles: don’t obsess over position. Their data showed a hospital that appeared in 69 of 71 ChatGPT answers but ranked first in only 25 of them, appearing consistently is the signal, and exact rank is mostly lottery. That also means you should be skeptical of any vendor selling precise “AI rankings”; Fishkin’s own conclusion was blunter than mine, and his was “full of baloney.”

And if you’re wondering whether this is worth your time at all: ChatGPT alone reported 800 million weekly users in October 2025, and SE Ranking’s 2026 traffic study put ChatGPT at roughly three-quarters of all AI referral traffic. People are asking. Somebody’s getting recommended.

The manual protocol, step by step

Five steps. One afternoon to set up, then about 30 minutes a month to maintain.

1. Build your prompt set (10–20 prompts, four types)

Write prompts the way a customer would talk, not the way a keyword tool would. Cover four types:

  • Category: “What’s a good CRM for a 10-person marketing agency?”
  • Comparison: “[Your product] vs [competitor] for [use case]”
  • Best-of: “Best accounting software for freelancers in 2026”
  • Problem: “What’s the easiest way to schedule social posts across five clients?”

If you’re local, add a “near me” variant. If you serve two very different niches, write a prompt set for each. Save the exact wording in a document, because the verbatim text is what you’ll rerun every month.

2. Run each prompt in a fresh session across all three assistants

New chat for every prompt. No memory, no carried-over context, no personalization from your account history. Use each assistant’s default model, and note which version it shows, because defaults change without announcements.

Keep the setup identical between months: same logged-out state, same platform (web or app), same country. Comparing a logged-in ChatGPT Plus session this month to an incognito run next month is comparing two different experiments.

ChatGPT, Gemini, and Perplexity are the core three. Add Google’s AI Overviews and Copilot later if you want extra columns, but don’t let scope creep stop you from starting.

3. Score mentions, citations, and rough position

A mention is your brand name appearing in the answer text, with or without a link. A citation is your website being linked as a source the model used to build the answer.

Score each run on three things: mentioned yes or no, cited yes or no, and where you appeared in any list (first, top three, buried, absent). Rough position is enough, remember, exact rank is noise.

Share of model is the percentage of AI answers in your category that mention your brand, think share of voice, but measured against model answers instead of search results. It’s the single number I’d trend month over month, and it’s what most tracking tools are really selling you under fancier names.

Mentions and citations come from different places, which matters when you’re diagnosing. Mentions reflect whether you exist in the model’s world: training data, reviews, directories, discussions. Citations reflect whether your pages get retrieved and judged worth quoting. You can absolutely have one without the other.

4. Log every source the assistants cite

This is the most underrated step. For each answer, note which domains got cited. Patterns show up fast: Wikipedia, Reddit, a few listicles, review platforms, industry news.

Citation studies through 2025 and 2026 keep finding Wikipedia and Reddit at or near the top across assistants, though the exact shares swing a lot by category. Perplexity typically attaches more sources per answer than ChatGPT or Gemini, so it exposes more of the source set for you to study.

If your site never gets cited but the same three listicles always do, you’ve just learned exactly where your gap is. That’s more actionable than any visibility score.

5. Repeat monthly, then compare like for like

Same prompts, same models, same fresh-session setup, roughly the same day each month. Compare against last month, not against one remembered answer.

One month’s dip is noise. Three consecutive months of dips is a trend worth investigating. And run ad-hoc checks after big changes: a new product launch, a major model update, or a competitor doing something loud.

What to record: the log template

Here’s the spreadsheet I’d set up. Seven columns, one row per prompt per assistant per run.

FieldWhat to record
DateWhen you ran the batch, keep it the same week each month
PromptThe exact text, verbatim from your fixed set
Assistant + modeChatGPT, Gemini, or Perplexity, note the model version if shown
Brands mentionedEvery brand in the answer, in the order it appeared
Your positionFirst mention, top 3, present, or absent
Cited sourcesEvery domain the answer links to
SentimentHow you were described: positive, neutral, negative, or wrong
NotesAnything odd: a new model, a much shorter list, a factual error

The sentiment column matters more than people expect. Being mentioned while being misdescribed is a problem to fix, not a win to celebrate. If an assistant confidently states you have features you don’t, log it, that’s usually a stale or wrong source shaping the answer.

Tools that automate the check

You don’t need a tool to start. But past ten prompts and three assistants, automation saves real hours.

ToolWhat it tracksModels coveredStarting price
Ahrefs AI Visibility CheckerOne-off snapshot: mentions, top cited domains and pages, triggering topicsChatGPT, Gemini, Perplexity, Copilot, AI Overviews, AI ModeFree, no signup (limited preview)
Semrush AI Visibility CheckerSnapshot: AI visibility score, mentions, citations, competitor comparisonChatGPT, Gemini, AI Overviews, AI ModeFree, 3 checks/day, no account
Semrush AI Visibility ToolkitDaily tracking on 25 custom prompts, competitors, prompt research, site auditChatGPT, Google AI, Gemini, Perplexity$99/mo per domain, billed annually; 7-day trial
Ahrefs Brand RadarOngoing tracking, AI share of voice, estimated impressions, 519M-prompt indexAI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot; Claude via custom promptsCustom prompts from $50/mo; standalone index from $199/mo
Otterly.AIDaily prompt tracking, sentiment, citation source tracking, GEO auditsChatGPT, AI Overviews, Perplexity, Copilot; Gemini, AI Mode, and Claude as paid add-ons$29/mo for 15 prompts
Peec AIVisibility, sentiment, competitor benchmarking, source citationsChatGPT, AI Mode, AI Overviews, Copilot, Gemini; Perplexity and Claude on Enterprise only$95/mo for 50 prompts
ProfoundEnterprise tracking, agent analytics, shopping visibility, citation analyticsUp to 9 engines incl. ChatGPT, Perplexity, Gemini, Copilot, Claude, DeepSeekCustom pricing; free 7-day trial

A few notes before you pull out a credit card:

  • Check model coverage before price. Peec’s self-serve plans don’t include Perplexity until Enterprise, which is a real problem if Perplexity citations are your main concern. Otterly covers it on the cheapest tier.
  • Free checkers are snapshots, not trends. Ahrefs’ free tool shows a limited top-5 preview, and Semrush’s limits you to three checks a day. Great for a first look, useless for a monthly baseline.
  • Ask any vendor how they sample. After the SparkToro numbers, a tool that runs each prompt once and reports exact rank deserves your suspicion. Ask how many runs per prompt they do and how they aggregate.

What to do with the results

Three patterns cover most outcomes, and each has a different next move.

  1. You’re absent everywhere. Don’t panic, and don’t rush to buy a GEO package the same week. Start with the audit-first approach I’d take before spending anything on GEO, then work on the fundamentals that make a brand show up in ChatGPT recommendations at all.
  2. You’re mentioned but never cited. That means your product exists in the model’s world but your pages aren’t being retrieved or trusted. Study who does get cited, the listicles, review platforms, and Reddit threads in your log are your map. The usual fix is showing up where the answers already live, not publishing more thin pages.
  3. You’re cited but described wrong. Fix the facts on your own pages first, then the source the assistant pulled from. Stale pricing or outdated feature lists on a comparison site will keep misleading answers long after you’ve fixed your site.

The competitor angle deserves its own read, because when rivals show up and you don’t, the cited sources usually explain why.

FAQ

How do I check if ChatGPT recommends my business?

Ask 10–20 buyer-style prompts in fresh chats (category, comparison, best-of, and problem queries) and note whether your brand appears, where it ranks, and which sources get cited. For a five-minute snapshot instead, Ahrefs’ free AI Visibility Checker runs search-backed prompts for you with no signup.

How do I track brand mentions in AI answers?

Use the log template above in a spreadsheet, one row per prompt per assistant per run, and track mentions separately from citations. If that gets tedious, tools like Otterly ($29/mo) or Peec ($95/mo) automate the runs and give you share-of-model style trends without the copy-pasting.

Which tools track AI visibility?

Options as of October 2026: Otterly.AI from $29/mo, Peec AI from $95/mo, Semrush’s AI Visibility Toolkit at $99/mo per domain, Ahrefs Brand Radar from $50/mo for custom prompts or $199/mo for the standalone index, and Profound on custom enterprise pricing. Free snapshots: Ahrefs’ and Semrush’s free checkers.

How often should I test AI answers?

Monthly for a baseline trend, since AI answers change gradually enough that weekly checks add noise without much signal. Go weekly only during campaigns or big content pushes, and always add an ad-hoc run after a major model update or product launch. Consistency of setup matters more than frequency.

Why do AI answers change every time?

Because they’re generated fresh each run: models sample probabilities, fetch different live web results, and get quietly updated by their makers. SparkToro’s 2026 research found the same prompt run 100 times almost never produces the same brand list twice. That’s exactly why the protocol uses a fixed prompt set and repeated sampling.

Can I do this without paying for a tool?

Yes, that’s the point of the manual protocol. The spreadsheet version costs you 30 minutes a month and gives you a defensible baseline. Tools buy you scale and history, not different information.

Sources

Questions & answers

What is "How to Check Whether ChatGPT, Gemini, and Perplexity Recommend Your Business" about?

Run a free AI search visibility check on ChatGPT, Gemini, and Perplexity with a 30-minute monthly protocol, a log template, and tools worth paying for.

Who wrote this article?

Aditya Mallah is a growth marketer for SaaS, AI tools, and fintech. Full bio: https://adityamallah.com/about

About this article

The content on this blog is for educational and informational purposes. These are my own views and do not constitute professional, financial, legal, or business advice. I can be wrong; do your own research before making decisions. Third-party links are provided for convenience and do not imply endorsement.

Was this article helpful?

Get my growth notes

Notes on growth, AI, and finding customers. Delivered to your inbox.

All articles

B2B marketing 12 min read

If I Had to Grow a B2B Brand from Zero Today, Here's My 90-Day Plan

Here's the exact 90-day plan I'd use to grow a B2B brand from zero: fix positioning, pick one channel, build compounding assets, and run sales yourself.

keyword research 11 min read

AI Keyword Research Tools: Can They Replace Traditional SEO Platforms?

AI keyword research tools handle ideation and clustering well, but they can't produce real search volume. Here's the split that actually works in 2026.

GEO 12 min read

Best GEO Software for Brands: Comparing AI Visibility Tracking Platforms

Compare the best GEO software of 2026 by price, engine coverage, and limits, with picks for small budgets, mid-market teams, and enterprise brands.