Is Your Brand Actually Showing Up in ChatGPT, Perplexity, and Gemini — Or Did You Just Get Lucky Once?
Here’s how most teams “track” AI visibility today: someone types the brand name into ChatGPT, sees a mention, takes a screenshot, and drops it in Slack. Three weeks later, in a board meeting or a client call, the question comes back — “are we still showing up?” — and nobody actually knows. The one data point was never repeated, never logged against a prompt, and never compared across engines.
That’s not tracking. It’s a lucky glance. And it creates a real problem: you can’t prove the ROI of GEO/AEO work without a baseline, you can’t catch it when your brand silently drops out of an answer the following week, and you can’t tell a client or a founder anything more concrete than “I think we’re doing okay.”
Quick Answer
Track AI visibility by running the same 15–20 prompts (branded, category, problem-solution, comparison) across ChatGPT, Perplexity, and Gemini every 2 weeks, logging appearance rate, citation rate, position, and sentiment in a simple spreadsheet. The full template is below.
This post gives you the system we use internally — the exact prompt set, each platform’s quirks, the four metrics that actually matter, and a reporting template you can copy today. No paid tool required to start.
Build Your Prompt Set Before You Touch Any Tool
The mistake most people make is going tool-first — signing up for an AI-visibility platform before deciding what they’re actually going to ask it. Start with the questions real buyers would ask, not your own brand name repeated five ways. A solid starting set covers four categories:
- Branded — “What is [Your Brand]?”, “Is [Your Brand] good for [use case]?”
- Category — “Best [your category] for [your ICP]”
- Problem-solution (JTBD) — the actual job the buyer is trying to get done, e.g. “How do I stop losing customers to AI search results?” rather than “AI SEO services”
- Comparison — “[Your Brand] vs [Competitor]”, “Alternatives to [Competitor]”
15–20 prompts across these four buckets is enough to see a real pattern without drowning in data. Fewer than that and one answer skews your whole read; more than that and you won’t keep it up consistently, which defeats the point.
The 3 Platforms, One at a Time
This assumes you’ve already got some citation-building work underway — if not, start with how to get cited in ChatGPT and Google AI Overviews first, then come back to measure it. Run the identical prompt on all three and you’ll often get three different answers — that’s not a bug in your tracking, it’s how each one actually works.
ChatGPT
ChatGPT leans on a mix of training data and live browsing (when the query triggers a search). For tracking purposes, run every prompt twice — once plain, once with “search the web” implied by asking something time-sensitive (e.g. “in 2026”) — because the two modes can give different answers about the same brand. Log which mode produced which result; this is the single most-skipped step in manual tracking and the reason two people checking “the same thing” on different days get different answers.
- Run each prompt twice: plain, and with a time cue like “in 2026” to force live browsing mode
- Note whether the answer comes from training data alone or cites a live source link
- Check a fresh/incognito session separately from a logged-in one with memory enabled — custom instructions can skew results
- Flag outdated mentions — ChatGPT’s training cutoff can surface stale competitor rankings
Perplexity
Perplexity is citation-first — every answer comes with visible numbered source links, which makes it the easiest platform to audit for your actual citation rate. If your prompt set returns an answer that mentions your brand but the sources panel doesn’t link to anything on your domain, that’s a real signal: you’re being mentioned from secondary sources (reviews, directories, competitor comparison pages) rather than your own content, which is a different problem to fix than simply “not being mentioned” at all.
- Always check the numbered Sources panel, not just the answer text — that’s where your real citation rate lives
- Note whether the citation is your own domain vs. a directory or review site mentioning you
- Compare “Concise” vs “Detailed” answer modes — citation depth can change between them
- Re-check weekly while actively publishing new content, since Perplexity tends to recrawl faster than the other two
Gemini
Gemini pulls heavily from Google’s own index, so its answers often track closer to your existing organic search rankings than ChatGPT’s or Perplexity’s do. That makes it the best early-warning signal if your citation rate is dropping purely because your organic rankings slipped — check your Search Console data alongside Gemini results before assuming a GEO-specific problem when it might just be a ranking problem wearing an AI-search costume.
- Cross-reference results against your Search Console rankings for the same query
- Check both the AI Overview snippet (if triggered) and a direct Gemini app query — they don’t always agree
- Note competitors Gemini surfaces that ChatGPT/Perplexity don’t — often signals a different source like Google Business Profile or YouTube
- Re-check after any major Google algorithm update, since Gemini tracks closer to core search than the other two
The 4 Metrics That Actually Matter
| Metric | What it tells you |
|---|---|
| Appearance rate | Of your prompt set, what % actually mention your brand at all |
| Citation rate | Of the answers that mention you, how many link to your actual site/content |
| Position | Are you the first brand named, or buried after three competitors |
| Sentiment/framing | Are you described accurately and favorably, or vaguely/incorrectly |
Appearance rate and citation rate are the two to watch first — they’re the clearest signal of whether your GEO work is translating into actual presence, not just rankings on a page nobody reads anymore.
The Reporting Template (Copy This)
This is the part most guides skip. Below is the exact row structure to log every check — in a spreadsheet, it’s one row per prompt per engine per date:
| Date | Prompt | Engine | Appeared? | Position | Competitors Named | Source Cited | Sentiment |
|---|---|---|---|---|---|---|---|
| 2026-10-01 | “Best AI SEO agency for startups in India” | Perplexity | Yes | 2nd | Competitor A, Competitor B | Yes — blog post linked | Positive |
| 2026-10-01 | “Best AI SEO agency for startups in India” | ChatGPT | No | — | Competitor A, Competitor C | No | — |
| 2026-10-01 | “Best AI SEO agency for startups in India” | Gemini | Yes | 1st | Competitor A | Yes — case study linked | Positive |
(The rows above are a worked example to show the format — fill in your own results, don’t copy this data.)
Run the full prompt set across all three engines, log every result this way, and after 4–6 weeks you have something you’ve never had before: an actual trend line instead of a single screenshot.
How Often Should You Check?
Weekly is overkill for most brands and burns time for no extra signal. Every 2 weeks is enough to catch real drift during a stable period. If you’ve just published new GEO-focused content, shipped new schema, or are actively running a citation-building push, tighten that to weekly until you see the needle move — then pull back to the 2-week cadence. Once you’re checking weekly, doing it by hand gets tedious fast — our review of 10 AI SEO tools covers a few that automate this exact prompt-and-log workflow.
Don’t want to build and maintain the spreadsheet yourself? Get a free AI visibility check and we’ll run your first prompt set across all three engines for you.
What Changes Once You Actually Track This
The real shift isn’t the data itself — it’s what you can do once you have it. A rising citation rate is hard evidence a GEO investment is working, long before it shows up as a lead. A sudden drop in appearance rate after a competitor publishes a comparison page is an early warning, not a surprise three months later when the pipeline dries up. This is the same measurement discipline behind the client work in our case studies — the pattern holds across industries: you can’t manage what you don’t track, and AI search is no exception.
If this measurement gap is also why your CAC has been creeping up, tracking is the first step back to a real answer — not a guess.
Ready to See Where You Actually Stand?
Run the prompt set yourself this week, or skip the manual work entirely.
FAQ
What is AI visibility tracking?
AI visibility tracking is the practice of repeatedly running a consistent set of prompts through AI answer engines — ChatGPT, Perplexity, Gemini and similar — to measure whether, how, and how favorably your brand is mentioned, logged over time rather than checked once.
Which AI platform should I track first if I can only pick one?
Perplexity, because it’s citation-first and shows you directly whether your actual content is being referenced, not just whether your brand name is recognized.
How many prompts do I need to track AI visibility accurately?
15–20 prompts spread across branded, category, problem-solution, and comparison questions is enough to see a reliable pattern without the tracking becoming unsustainable to maintain.
Is there a free way to track ChatGPT visibility without buying a tool?
Yes — manually running your prompt set across each engine and logging the results in a spreadsheet using the template above costs nothing but time, and is a reasonable starting point before investing in a paid platform.
How often should I check my AI visibility?
Every two weeks during a stable period; weekly if you’ve just published new content or are actively working to build citations and want to see the impact sooner.
Is AI visibility the same as my Google ranking?
No. You can rank well in traditional Google search and still be invisible in ChatGPT or Perplexity’s answers, because these engines retrieve and synthesize differently than a classic search index — which is exactly why separate tracking matters.
