The LLM Citation Tracker

What this is — and why it matters

The one-sentence version: AI answer engines now hand customers two or three business names instead of ten blue links — and if yours isn’t one of them, you lost the customer before they knew you existed. This tracker is the cheapest, most honest way to find out whether you’re on that shortlist, starting this week, with nothing but a spreadsheet and ten minutes.

This document describes a manual methodology for tracking when ChatGPT, Claude, Perplexity, Gemini, and Google’s AI Overviews cite your business or your published content in their answers. It also provides a spreadsheet template you can copy and begin using this week.

The methodology takes roughly 10 minutes per week. It requires no tools, no subscriptions, and no technical setup. It gives you the most direct feedback available on whether your GEO efforts are producing results. Most businesses have no idea whether they’re being cited or not — only around 16% of brands systematically track AI search performance (HubSpot, citing Sep 2025 data). You’re about to be in the minority that actually knows.

The honest caveat

There is no public API that tells you when an AI names your business. ChatGPT does not notify you. Perplexity sends no alerts. The only way to know is to ask the engines yourself, manually, using the questions your customers would actually type. Running 10 queries across four AI platforms takes about 10 minutes. Do it once a week, record the results in the tracker, and review the pattern after four weeks.

Real limits to read your data correctly:

• You only see the queries you think to run — test the questions customers actually ask, not your business name.
• Answers vary run to run. Ask ChatGPT the same thing 100 times and it surfaces ~44 different brands, but only ~5 appear 80%+ of the time (Search Engine Land / Fishkin, Feb 2026). The trend over weeks is the signal.
• Log out and use a fresh/incognito session so results are not personalised to you.
• "Mentioned" and "cited with a link" are different things. Track both.

This is not a perfect instrument. It is the best free, honest, do-it-yourself instrument that exists — and it gives you genuine, unmediated feedback no dashboard can fake. Run it with discipline and the pattern becomes undeniable within a month.

The engines — and why you test all of them

Each engine pulls from different places and updates on its own clock. Being named in one tells you little about the others. SOCi’s 2026 Local Visibility Index found only 45% of the top-20 traditional local-search brands also appear in the top-20 AI recommendations (SOCi, ~350,000 locations). Different engines, different doors, different winners.

What to record — 5 signals per query

The weekly routine — 10 minutes, same day, every week

Weekly is the right rhythm. Retrieval indexes change constantly — in one study, Reddit’s share of ChatGPT citations swung from ~60% of responses to ~10% inside six weeks (Semrush, 230k+ prompts). Monthly testing is too coarse to act on; daily testing is noise.

1. Open a fresh/incognito browser session (so results are not personalised to you). Log out of AI accounts where you can.
2. Run your live queries across the engines that matter to your audience — a sustainable default is your top 8–10 queries across ChatGPT + Perplexity + Gemini, rotating Claude and Google AI Overview in.
3. Log each result as one row in the Weekly Log: query, engine, Cited Y/N, position, source URL, sentiment, competitor cited, a one-line note.
4. Glance at the Dashboard. Your citation rate, per-engine breakdown, share-of-voice vs competitors, and sentiment mix update automatically.
5. Done. Close the laptop. Ten minutes.

After four weeks, you have a trend you can trust. A realistic, healthy pattern: nothing for the first couple of weeks, then a first appearance once your content gets indexed and trusted — climbing from “none” to “list” to a named position.

When to automate — and where Ozvor fits

This spreadsheet is genuinely useful, and for a single-location business testing a handful of queries it may be all you need for a while. But it has a ceiling you’ll hit:

• Coverage: 10 minutes covers ~10 queries. Real categories have hundreds of phrasings.
• Memory and consistency: one person eyeballing answers is subject to mood, fatigue, and missed incognito sessions.
• Competitors and benchmarking: tracking share-of-voice by hand across multiple rivals and engines gets unmanageable fast.
• Scoring and content: the spreadsheet tells you that you're invisible. It doesn't tell you your overall standing or exactly what to publish next.

That’s the line where you graduate to Ozvor — the AI Search Trust Intelligence platform this tracker comes from. Ozvor automates everything above, computes your TrustIndex Score (Brand 30% / Performance 35% / AI 35%), benchmarks you against your competitors, and builds a GEO content plan so you’re not guessing what to publish.

Included with paid plans — or automate it entirely

The tracker and methodology are free to download. When you’re ready to scale:

• Free — 1 brand, 3 competitors, 50 prompts, monthly audit and TrustIndex Score. No credit card.
• Growth — $99/mo (or $831/yr with founder pricing) — 1 brand, 10 competitors, 250 prompts, weekly automated monitoring, citation tracking, and GEO content. This is the automated version of this spreadsheet at 25× the coverage.
• Agency — $149/mo (or $1,251/yr founder) — up to 25 brands, white-label reports, client workflow.