Guide

How to track brand mentions in AI search

A practical method for tracking whether ChatGPT, Perplexity and Google AI Mode mention your brand. Manual prompts, metrics, tools and a monthly cadence.

By Sunny Patel · Updated 2026-08-23

AI assistants now answer buying questions directly. A handful of brands get named when someone asks ChatGPT for the best accounting software for freelancers. Everyone else is invisible. This guide shows you how to find out which group you are in, first by hand and then with tools. It also shows you how to turn the answer into a simple monthly routine.

This guide is about tracking your brand inside AI assistant answers, in ChatGPT, Perplexity, Claude, Gemini and Google AI Mode. Tracking mentions across news sites, blogs and social media is traditional media monitoring. Use a tool like Google Alerts or a social listening platform for that job instead. The two overlap but they are not the same job.

The manual method: ask the assistants yourself

You do not need a tool to get a useful baseline. You need an hour and a spreadsheet.

Start by writing 10 to 20 questions a real buyer in your niche would ask. Focus on buying intent rather than brand searches. Nobody needs a tool to confirm that asking an AI about your brand returns your brand. The questions that matter are the category ones: best X for Y, X vs Z, is X worth it, alternatives to the market leader, how to choose an X.

Ask each question in ChatGPT, Perplexity, Claude and Google AI Mode. Record what comes back. Note whether your brand appeared at all, where it appeared in the answer, what the assistant said about you, and which websites it cited.

Three rules make the results honest. First, use fresh sessions. Assistants use earlier messages in a conversation as context. A later mention proves nothing if you already mentioned your brand earlier in the chat. Open a new chat for every question. Use a logged-out or incognito session where you can. Memory features can personalise answers towards brands you already use.

Second, vary your phrasing. Ask the same underlying question three or four different ways. Best CRM for small business, which CRM should a five person company use, and affordable CRM recommendations can produce different brand lists.

Third, accept that results are non-deterministic. Large language models sample from probabilities rather than looking up a fixed result. Many answers also depend on live web retrieval that changes day to day. The same prompt can name you on Monday and skip you on Tuesday. A single check is a snapshot rather than a verdict. You are measuring a rate over time: how often you appear across a set of prompts. A single appearance does not count.

What to measure

Four things are worth recording for every prompt and platform.

• Presence. Did your brand appear in the answer at all? This is your core metric, tracked as a percentage across your prompt set.
• Position. Were you the first recommendation, one of five, or a footnote? Being named first in a shortlist is worth far more than a passing mention at the end.
• Sentiment. What did the assistant actually say? Recommended for beginners is different from a budget option with limited support. Note the framing in a few words.
• Citations. Which sources did the answer link or reference? This is the most actionable column. A comparison article that omits you but keeps getting cited is your outreach target. Citations to your own site show you which pages are earning the mention.

Automating it with tracking tools

Manual checking works for a baseline. It does not scale past a few dozen prompts. Repeated sampling matters because answers vary. Software handles that repeated sampling. A few established options:

Otterly.AI monitors brand mentions, citations and sentiment across ChatGPT, Perplexity and Google AI Overviews. It sits at the affordable end of the market. Peec AI covers similar ground with competitor comparison and sentiment tracking. Profound is an enterprise platform that tracks answers across ChatGPT, Perplexity, Microsoft Copilot and Google AI Overviews against real user prompt data. Ahrefs Brand Radar is bundled into existing Ahrefs plans. It tracks how often your brand appears in AI answers and which URLs the mentions were pulled from. This is convenient if you already pay for Ahrefs.

All of them do broadly the same thing: run your prompt set on a schedule, count appearances, and chart the trend. We compare these tools in more depth in our guide to AI search visibility tools at /learn/ai-search-visibility-tools. The honest advice for most small teams is to start manual, prove the habit, then pay for automation once the spreadsheet becomes tedious.

The prerequisite most people skip: can AI read your site at all?

Tracking tells you whether you are mentioned. It cannot fix the most common reason you are not: the AI systems cannot read your site. Assistants that browse the web rely on search-index crawlers such as OAI-SearchBot, Claude-SearchBot and PerplexityBot. Many sites block these crawlers in robots.txt without realising it. A CDN default, a security plugin or an old blanket rule is usually the cause.

Check three things before you invest in tracking. First, open yourdomain.com/robots.txt and look for rules that disallow OAI-SearchBot, Claude-SearchBot or PerplexityBot. Those govern answer visibility. GPTBot, ClaudeBot and Google-Extended control model training instead. Second, check that your key pages have real text content in the HTML rather than content that only appears after JavaScript runs. Some AI crawlers do not execute JavaScript reliably. Third, consider adding an llms.txt file, an emerging convention that gives AI systems a plain-text summary of what your site covers. It is not a ranking lever but it costs nothing and removes ambiguity.

The free sitemap.digital scanner checks all three in one pass. It reads your robots.txt permissions for the major AI crawlers, scores every page for AI crawlability, flags llms.txt presence, orphan pages and broken links, and generates ready-to-use fix files including an llms.txt and robots.txt snippets. No signup needed. Fix blocked crawlers first when the scan flags them. Tracking a brand that AI systems cannot read is measuring a locked door.

A monthly tracking cadence you can start this week

This routine takes about an hour a month once it is set up.

Week one, build the asset. Write 15 buying-intent prompts for your niche and put them in a spreadsheet with these columns: date, prompt, platform, mentioned (yes or no), position (first, listed, footnote, absent), sentiment note, sources cited, and competitor mentioned. That last column matters because the gap between your mention rate and your nearest competitor's is often more motivating than your own number.

Then run the first pass. Ask every prompt in at least two platforms with a fresh session each time. Fill in the rows. This is your baseline.

Rerun the same 15 prompts on the same platforms every month. Add new rows rather than overwriting old ones. Monthly is frequent enough. AI answer patterns shift over weeks rather than hours. Daily manual checking just adds noise to a non-deterministic signal. Glance at your citations column alongside the rerun. Getting included on a third-party page that keeps earning mentions in your category is probably your highest-leverage action for the month.

Rescan your site's crawlability every quarter. A plugin update or CDN change can silently reintroduce a block. Review whether your prompt set still matches what buyers ask. That is the whole system. This system is unglamorous. It fits in a spreadsheet. It is more than most of your competitors are doing.

Frequently asked questions