How to Track Brand Mentions in Claude and Gemini
Track brand mentions in Claude and Gemini using a fixed list of buyer prompts, fresh sessions, saved answers, and a simple mention log. Compare results to your site and fix gaps with cite-ready pages.
How to track brand mentions in Claude and Gemini: write ten to twenty buyer prompts your team already hears, run each prompt in a fresh Claude session and a fresh Gemini session with the same date stamp, log whether your brand is named, whether your domain is cited, and which competitors appear, then fix the missing page before you change the prompt.
When the fix belongs on a live site, run the AI brand visibility checker and read our AI SEO for B2B services firms guide for the page types that vertical needs.
ChatGPT checks alone miss half the buyer journey. Anthropic and Google ship separate assistants with separate retrieval. Anthropic documents Claude as a general assistant that can browse when enabled. Google documents Gemini as an assistant that can use Google Search when you turn search on. Neither product publishes a share-of-voice score you can export, so your log is the source of truth.
Why Claude and Gemini need their own log
Assistants pick answers from what they can retrieve and what already ranks as a clear block on the open web. A mention in one product does not copy to another. Treat Claude and Gemini as two channels in the same spreadsheet, not as duplicates of a ChatGPT column.
If you already audit ChatGPT, keep that workflow. Add columns for Claude and Gemini instead of combining scores. The GEO tracking tools guide covers products that automate multi-engine reporting when manual logs outgrow the team. When leadership wants a percentage from the same prompt set, read the share-of-voice in ChatGPT and Perplexity guide after your log has a month of rows.
1. Lock a buyer prompt list
Pull prompts from the same places you would for classic SEO:
- Discovery calls and demo notes.
- Support tickets that start with "which tool should we use."
- Search Console queries with impressions on your category, even when clicks are low.
Write each prompt exactly as a buyer would type it. Avoid internal codenames. One prompt equals one row. When two wordings mean the same job, pick one keeper wording so your trend line stays honest.
2. Run fresh sessions in each product
Old threads bias the model toward earlier answers. For each monthly check:
- Open a new Claude chat and a new Gemini chat.
- Use the same account tier your buyers use when possible.
- Enable browsing or search when that is how your category is researched.
- Paste the prompt once, without leading the model toward your brand.
- Save the full answer text and screenshot the citation panel if links appear.
Record the date, product, mode (search on or off), and prompt ID. That metadata matters when an answer flips after a model update.
3. Score mentions and citations
Use the same four codes in both columns:
| Code | What you saw |
|---|---|
| M+ | Brand named in the answer body |
| C+ | Your domain linked or footnoted |
| Comp | Named competitor without you |
| Skip | Neither you nor a direct competitor named |
Do not treat Skip as a win. It usually means the answer stayed generic. Comp is the row that should drive your content plan.
4. Fix the page gap, not the prompt
When Comp or Skip repeats, ask which URL should win the prompt. Often the fix is a comparison or how-to page with a direct answer in the first paragraph, not a longer chat transcript.
Ship or rewrite that page in crawlable HTML, with sources next to factual claims. The comparison brief guide shows how to brief those pages. The Google AI Overviews playbook shares block structure that also helps assistants quote you cleanly.
Run the free AI brand visibility checker when you want a structured visibility read on priority prompts before you commit to a production queue. When you are ready for sample pages in your design, start the free preview on your domain.
5. Recheck on a schedule
Monthly is enough for most teams until a launch or pricing change. Re-run the full prompt set, not only the rows that failed last time, so you notice regressions on prompts you thought were fixed.
Orchory runs the page side when mentions lag: buyer questions become cite-ready pages in your site design, and you approve before anything goes live. Start with the AI brand visibility checker on ChatGPT, Perplexity, and Google AI Overviews, or send your site for a first page on a call. Ongoing production is $990/mo; cancel anytime. Browse the free SEO tools hub for technical checks while you ship.
