Tracking brand mentions in AI means measuring how often, where, and how your brand is named in answers from ChatGPT, Claude, Perplexity, Gemini, and the AI Overviews layered on Google and Bing. It's the AI-era equivalent of media monitoring, with one key difference: there is no API, no Search Console, and no real-time feed. You have to manufacture the measurement.

This guide explains the workflow — what to track, how to sample, how to parse, and how to act on the data.

What "AI brand mention" actually means

For each AI surface, a brand mention is any of:

1. Direct mention — your brand name appears in the answer's plain text.
2. Cited mention — your domain is linked or attributed as a source.
3. Implicit mention — the answer describes your product without naming it (rare, but worth tagging).

The two metrics worth tracking weekly:

• Mention rate — % of relevant prompts where your brand is mentioned.
• Citation rate — % of relevant prompts where your domain is linked.

Mention without citation is worth roughly half as much as mention with citation, because the second drives traffic. Track both.

The four AI surfaces to monitor

If you're new to this and want one surface to start, pick ChatGPT. If you want two, add Perplexity. Three: add Claude. Four: add Gemini.

How to sample correctly

AI surfaces produce different answers run-to-run. The same prompt to ChatGPT can produce three different answers in three minutes. Single-sample monitoring is noise, not measurement.

The bar:

• 3+ samples per prompt per week per surface. Anything below 3 produces false confidence.
• Same time of day, same account context. Reduces variance from training-data updates or A/B tests.
• Don't paste your brand name into the prompt. Defeats the purpose; you want to know whether the model names you unprompted.

Build your prompt universe

A prompt universe is the set of buyer questions you'll monitor against. Build it in three layers:

• Awareness (40%): "What is [category]?", "Best [category] for [use case]"
• Consideration (40%): "[Competitor] vs [Competitor]", "Alternatives to [Competitor]"
• Decision (20%): "Is [your brand] good?", "[Your brand] reviews"

100 prompts is the minimum. 250+ is the comfortable floor. Below 100, your monitoring is too narrow.

The manual workflow (first 30–60 days)

Without a tool, the workflow is:

1. Open ChatGPT, Claude, Perplexity, Gemini in tabs.
2. Run each prompt 3 times per surface, weekly.
3. In a spreadsheet, log: surface, prompt, mentioned (y/n), cited (y/n), competitors named, citation URL.
4. Compute weekly aggregates.
5. Watch trends week-over-week.

This is sustainable for ~50 prompts × 4 surfaces × 3 samples = 600 runs/week. Beyond that, the manual cost exceeds any tool's price.

The tooled workflow

For continuous tracking at scale, Tracemetry Pro at $199/mo automates all of this:

• 250 custom prompts × 4 surfaces × 3 samples weekly = ~3,000 runs/week, automated.
• Per-prompt mention/citation parsing.
• Share of voice vs your direct competitors.
• Weekly digest email with the top mention-rate movers.
• Source-grounded brief generation for the gaps.

Other credible options: AthenaHQ ($300/mo, two surfaces), Profound (~$2,000+/mo, three surfaces, enterprise), Peec AI ($200/mo, four surfaces, EU-focused). See our full comparison of AI brand visibility tools.

What to do with the data

Tracking is half the work. The other half is acting on the data.

Weekly digest cadence

For every weekly run, compute:

• Mention rate per surface. Spot per-surface anomalies. Drops on one surface usually mean you've lost a competitor comparison or your page got stale.
• New competitors detected. Brands appearing in your prompts for the first time. Investigate why they're winning — usually a recent content push.
• Newly-lost prompts. You appeared last week, you don't now. These are the highest-priority gaps to close.
• Citation churn. Pages that used to be cited but aren't anymore. Usually a freshness problem.

Convert insights to work

For each newly-lost prompt:

1. Pull the current answer text. What does the model say instead?
2. Identify which brand is winning that prompt.
3. Read their page. What is it doing better?
4. Ship your own page or update an existing one. Use the content shape that wins.

For each newly-detected competitor:

1. Search their domain for the prompt's keywords.
2. Note their content shape, schema, and recency.
3. Decide whether to compete head-on or differentiate.

How a mention-rate drop typically plays out (a hypothetical example)

Picture a project-management SaaS whose ChatGPT mention rate drops from 28% to 19% in one week. Their pricing page had been updated, but the datePublished was bumped to a fresh date while the underlying content was unchanged. ChatGPT's freshness signal picks up the new date; the model retrieves the page; it finds nothing materially new; and it switches to citing a fresher competitor.

The fix is straightforward — rewrite the pricing FAQ block with genuinely new examples, then keep the new date. The recovery curve usually trails one week behind the rewrite.

Without weekly tracking, this drop is invisible. A team would notice pipeline softening three months later and never connect it to AI visibility. The cost of not monitoring is far larger than the cost of monitoring — which is the entire argument for the discipline. Google's own guidance on keeping content fresh treats freshness as a real signal; AI engines weigh it even more heavily.

Common tracking mistakes

• Setting up monitoring once, never reviewing. Tracking without weekly review is theater.
• Trusting single samples. 30–50% variance run-to-run. One run is noise.
• Tracking your brand only. Track competitors too. Their wins are your gaps.
• Mention-only, no citation. Citation drives traffic. Mention without citation is half the value.
• No prompt universe. Random ad-hoc prompts don't aggregate. You need a defined set you re-run weekly.

FAQ

How do I track brand mentions in ChatGPT? Define a set of relevant prompts, run each 3+ times per week, parse the answers for your brand name. Either manually (spreadsheet, ~30 prompts) or with a tool. Tracemetry automates this across ChatGPT, Claude, Perplexity, and Gemini.

Can I track mentions in Perplexity? Yes — Perplexity is the easiest AI surface to track because almost every answer cites its sources directly. Manual tracking works well for the first 30 prompts; beyond that, use a tool.

What's the cheapest way to track AI brand mentions? Tracemetry Tracker at $39/mo for the cheapest tier with credible multi-surface coverage. Otterly's free tier for very light tracking on two surfaces. Manual spreadsheet tracking for under 30 prompts.

How often should I track? Weekly. AI surfaces drift fast enough that monthly tracking misses meaningful changes. Quarterly is too slow to act on. Daily is unnecessary noise.

What if my brand has no mentions yet? That's the starting point for ~30% of our customers. The work is building the content and authority signals that earn the first mentions. Run the free audit to see exactly which prompts in your category have no current mention of your brand, and what would change that.

Start with the audit

The fastest baseline: free public audit. Three prompts across ChatGPT, Claude, and Perplexity, no signup, results in 60 seconds. Shows your current mention rate, top competitors named instead, and three gaps to close.

For continuous weekly tracking across 250 prompts and four surfaces, Tracemetry Pro at $199/mo.