AI search

How to track your brand's visibility in ChatGPT

Your buyers now ask ChatGPT what to buy. Here's how to track your brand's visibility in its answers, by hand or with your agent: which prompts surface you, which competitors it recommends instead, which sources shape the answers, and what to fix.

Jesse Sibley
By , FounderPublished · Last updated · 12 min read

Why ChatGPT visibility doesn't show up in your analytics

When someone asks ChatGPT "what's the best CRM for a five-person startup", they get a short list of recommendations, not ten blue links to explore. If your brand isn't on that list, you're invisible to that buyer, and traditional analytics won't tell you. The only way to measure your AI visibility is to ask the engines yourself, the way your customers do, and track the answers over time.

AI answers differ by engine and prompt phrasing. ChatGPT might recommend your product for "best project management tool for remote teams" but not for "project management software for startups". Gemini and Claude read different sources and make different recommendations. What worked last month might not work today, because the engines update constantly and the sources they cite shift.

The AI visibility loop: write prompts, ask the engines, log the answers, fix the sources, then re-check
The five steps below, run again each month.

Step 1: Write the prompts your buyers actually ask

Start with the language your customers use in sales calls, support tickets, and community threads like Reddit or forums in your category. Avoid SEO keywords: AI engines respond to natural questions, not keyword strings.

Good prompts are specific and often comparative:

  • "best CRM for startups under 10 people"
  • "HubSpot alternatives for a small marketing team"
  • "is [your brand] good for agencies?"
  • "Salesforce vs [your competitor] for enterprise"

The same works for a consumer brand. A clothing label might track "best sustainable sneakers under $150", "brands like Everlane but cheaper" and "is [your brand] true to size?".

Group your prompts by buyer intent:

  • Category discovery: "best X for Y" questions where buyers are learning about the space
  • Direct comparisons: "A vs B" or "alternatives to X" when they're narrowing options
  • Brand research: "is [brand] good for [use case]" when they're validating a choice

Each group reveals different competitive gaps. Category questions show whether you're in the consideration set at all. Comparisons show who you're up against. Brand research questions reveal whether the engines understand your positioning.

Collect 20 to 40 prompts to start, enough to cover the main paths through your buyer's journey. You'll add more as new questions emerge from customer conversations.

Step 2: How to check what ChatGPT, Gemini and Claude recommend

The engines don't agree on recommendations, and they read different sources at different times. Some answer from what the model already knows unless they search the web for that question.

Test your prompts on these engines:

EngineWhere to askNotes
ChatGPTchatgpt.comStart here: it is the one most buyers know
Geminigemini.google.comGoogle's assistant, separate from Search
Claudeclaude.ai, with web search onPopular with technical buyers
Perplexityperplexity.aiSearches the web for every answer
Google AI OverviewsGoogle SearchThe AI summary above the results, when Google shows one
Google AI ModeThe AI Mode tab in Google SearchGoogle's conversational search

For each prompt-engine pair, run it fresh in a new chat, logged out or in a private window, to avoid personalization. Note the exact date: answers can shift within days as the engines reindex sources or update their models.

You don't need to test every engine every time. Start with ChatGPT and Gemini, then add others based on where your buyers spend time.

Step 3: Is there a way to see if ChatGPT and Claude recommend your competitors instead of you?

Yes. For each AI answer you collect, log what it said: whether your brand was mentioned and in what position, which competitors appeared, whether it described you correctly, and which sources the engine cited. That sheet shows which engines recommend you, who you are up against in AI search, and the pages shaping your visibility.

Log these for each answer:

  1. Was your brand mentioned? Note yes or no and its position in the list (1st, 3rd, not named).
  2. Which competitors were named? List them in the order shown. This reveals who you're really competing against in AI answers, which might differ from your assumptions.
  3. Was the description right? Note anything wrong or out of date: pricing, features, who you're for.
  4. Which sources were cited? Copy the URLs the engine linked to. These are the pages shaping your AI search visibility.

A simple log sheet looks like this (the rows are an illustration, not a real run):

DatePromptEngineYour brand?PositionDescribed correctly?Competitors namedCited sources
2026-09-30best CRM for startupsChatGPTYes3rdYesHubSpot, Pipedrive, Salesforceg2.com/best-crm, productboard.com/blog/crm-startups
2026-09-30best CRM for startupsGeminiNoNot namedn/aHubSpot, Zoho, Mondaycapterra.com/crm, forbes.com/advisor/crm
2026-09-30HubSpot alternatives for small teamsChatGPTYes2ndOld pricingPipedrive, Zoho, Freshsalesreddit.com/r/sales, saashub.com/hubspot-alternatives
2026-09-30HubSpot alternatives for small teamsClaudeNoNot namedn/aPipedrive, Close, Copperalternativeto.net/hubspot, getapp.com/alternatives

In this example, the brand appears for the broad category question on ChatGPT but not Gemini, and shows up in the alternatives question on ChatGPT but not Claude, with its old pricing. The cited sources vary widely: comparison sites, a Reddit thread, review platforms, and a Forbes roundup.

Here is a real run of the same prompt, "best CRM for startups", in four engines on 6 October 2026. Each engine names a different set of brands, in a different order.

Chart of the CRMs four AI engines named for best CRM for startups, with HubSpot first in all four
Asked from the United States on 6 October 2026 with Manifold's aeo_run_ai_answers. Claude's engine did not answer that day. Folk, Freshsales and Copper were checked and never named.

This log reveals three immediate insights:

  1. Your AI visibility depends heavily on which engine someone uses
  2. Different prompt phrasings produce different competitive sets
  3. The sources shaping your visibility are scattered across the web, and you don't control most of them

How to measure your visibility in ChatGPT: four numbers

A log is easier to act on once you turn it into a few numbers you can compare month to month. Work them out per engine, and across all engines:

  • Mention rate: answers that name you, divided by answers checked.
  • Share of voice: your mentions, divided by every mention of a brand you track in the same answers.
  • Average position: your average place in the answers that name you.
  • Citation rate: answers that link a page on your own site, divided by answers checked.

In the real run above, Pipedrive was named in three of the four answers, a 75% mention rate, and was third each time. Across the four answers the engines named tracked brands 19 times, and HubSpot took 4 of those mentions, a share of voice of about 21%. With 20 to 40 prompts, the same arithmetic tells you which engines and which kinds of question you're losing.

Keep the "described correctly" column next to these numbers. Being named with the wrong price can cost you more than not being named.

Engines that search the web build their recommendations largely from the pages they cite, which are usually comparison sites, review platforms and editorial roundups you don't control. If those pages leave you out or describe your product incorrectly, the engines are unlikely to recommend you, even if your own site ranks well for the same query.

The cited sources tell you exactly where the engines learn about your category. Read each one and ask:

  • Does this source mention you? If a comparison chart or roundup appears in multiple answers but leaves you out, that's a pitch opportunity.
  • Is your own content cited? If your homepage, docs, or blog appear, check that they position you clearly for the use cases buyers ask about.
  • Are the facts correct? AI engines sometimes misinterpret sources or mix details across competitors. If a source gets your pricing, features, or positioning wrong, contact the author.

Four priorities follow:

1. Improve your cited pages. If the engines link to your homepage, case study, or blog, make sure those pages clearly explain who you're for and what problems you solve. Add specific use cases, comparison tables, and customer outcomes. The engines extract recommendations from detail, not marketing copy.

2. Get into the cited comparison sources. If G2 comparisons, Forbes roundups, or Reddit threads appear repeatedly in citations, those are the pages shaping your AI visibility. For review platforms, claim your listing and add detail. For editorial lists, pitch the authors with specific data on why you belong.

3. Let the search crawlers in. Check that robots.txt doesn't block OAI-SearchBot, the crawler ChatGPT search uses; OpenAI says sites that block it won't appear in ChatGPT search answers. For Claude, the equivalents are Claude-SearchBot and Claude-User. GPTBot, ClaudeBot and Google-Extended control model training, not whether you can be cited in search, so blocking those is a separate decision. Our OAI-SearchBot guide has the robots.txt lines for each.

4. Make sure Bing has your pages. OpenAI says ChatGPT search works with third-party search providers, Bing among them, so a page Bing hasn't indexed may be harder for it to find. Check your key pages in Bing Webmaster Tools, or ask your agent through a Bing Webmaster Tools MCP server.

For what ChatGPT actually cited across six buyer questions, and how to get onto those pages, see how to get cited by ChatGPT.

Also cross-check the sources competitors are cited from. If a rival appears in five answers and four cite the same guide or comparison post, that source is worth more attention than one that appears once.

Step 5: Re-check monthly to catch changes

AI answers aren't static. The engines update their models, reindex sources, and adjust recommendations constantly. A prompt that names you today might not next week, especially if a new comparison post publishes or a competitor gets featured in a high-authority source.

Run the same prompt set monthly as a baseline. Our AI Recommendation Index does exactly this for marketing MCP servers: 33 fixed prompts on five engines each month, with every answer published. For prompts that drive the most revenue, questions early in the buyer journey like "best [category] for [use case]", check weekly. To have Claude or ChatGPT run that weekly check through Manifold MCP, see how to track AI visibility with Claude or ChatGPT. Compare each run to the last and flag:

  • Prompts where you dropped out of an answer you used to appear in
  • New competitors who weren't named before
  • New cited sources that just started appearing

When you drop out, look first at what changed in the sources the engine cited. Did a new comparison chart publish? Did a cited source update and remove you? Did your own page stop being cited?

Track these changes in a second tab of your log sheet or a separate change log. Note the date, the prompt, the engine, what changed, and your hypothesis on why. This history reveals patterns: you might notice that you drop out of answers whenever a certain review site updates, or that you gain visibility after publishing certain types of content.

Automate AI visibility tracking with Manifold

The manual process above works, but it's time-consuming and hard to scale beyond a few dozen prompts. You're running queries by hand, copying answers into spreadsheets, and manually comparing runs month over month.

Manifold's AI visibility tools do the asking and logging for you. They run your prompt set across ChatGPT, Claude, Gemini, Perplexity and Google's AI answers whenever your agent asks, detect your brand and competitors in every answer, and extract cited sources. Instead of spending hours each month running prompts by hand, you get structured data showing which prompts surface your brand and which don't, and your agent can compare it with your last run.

For teams that want to connect their AI agent to AI visibility tracking, read our guide to the best AI visibility tools with an MCP server, which compares Peec AI, OtterlyAI, LLM Pulse, Ahrefs, Profound, Scrunch and others.

The tools return:

  • Whether each engine mentioned your brand, your competitors, or neither
  • The position of each brand in the answer (1st, 3rd, not mentioned)
  • Every URL the engine cited as a source

You can check hundreds of prompts and compare AI visibility by region or language. Run them once when you ask, or put the same questions in an AI visibility monitor, which asks them on a schedule and reports your share of answers against competitors as it moves. For teams managing AI search visibility at scale, this turns a manual research task into one request to your agent.

Read more in the AEO tools documentation, or try the tools directly if you already use Manifold.

Run this in your agent

Connect Manifold to Claude, ChatGPT or Cursor, then paste this prompt.

Check whether ChatGPT, Claude and Gemini recommend us for the questions our buyers ask. List who they name instead and which pages they cite, then tell me the three pages to fix first.

Questions people ask

Can I see how ChatGPT and Claude recommend my competitors instead of us?

Yes. Ask each engine the questions your buyers ask, in a fresh chat, and log which brands it names, in what order, and which sources it cites. The prompts where competitors appear and you don't are your gaps. Tracking tools, or your agent with an AI visibility MCP server, can run the same prompts for you.

Why is my brand never mentioned when people ask AI for product recommendations?

Usually because the pages the engines cite, such as comparison sites, review platforms, roundups and Reddit threads, leave you out or describe you poorly. It can also be that ChatGPT's search crawler, OAI-SearchBot, is blocked in your robots.txt. Read the cited sources first: they show where the engines learn about your category.

Is there a free way to track brand visibility in ChatGPT?

Yes. Write 20 to 40 buyer questions, ask them in ChatGPT and the other engines each month, and log the answers in a spreadsheet: whether you're named and where, which competitors appear, and which sources are cited. Paid tools save the time, not the method.

Can I see ChatGPT traffic in Google Analytics?

Partly. Visits from links in ChatGPT answers show up in GA4 as referral traffic from chatgpt.com. That tells you about clicks, not about the answers that named you without a click, so keep the prompt log as well.

How do I get ChatGPT to recommend my brand?

Get into the sources it already cites for your category, make your own pages state plainly who you're for and what you cost, keep OAI-SearchBot allowed in robots.txt, and make sure Bing has indexed your key pages. Then re-run your prompts monthly to see what moved.

Do AI answers change between runs?

Yes, frequently. The same prompt can name different brands week to week as engines update their models and reindex sources. That's why monthly tracking matters.

Jesse Sibley

Jesse Sibley

Founder of Manifold. A growth engineer who has worked with London startups for eight years.

  • 9 min read

    How to track AI visibility with Claude or ChatGPT

    Connect Claude or ChatGPT to Manifold MCP, create a weekly AI visibility monitor, and have your agent read share of answers, citations and what changed.

  • 17 min read

    The AI Recommendation Index for marketing MCP servers

    Each month we ask five AI engines the questions buyers ask about MCP servers for SEO, Search Console, Analytics, AI visibility, Reddit, LinkedIn, social, leads and ads, and count which brands they name and which sites they cite. Here are the leaders by category, the sources behind them, and the data.