Guide · Brand monitoring
How to track brand mentions in AI answers
AI surfaces like ChatGPT, Gemini, Claude, Perplexity, Grok, Google AI Overviews and AI Mode can recommend your brand or leave it out. This guide shows how to monitor your brand in each one by hand, how to keep results comparable, and when a tool is worth it.
7-day Base trial · No card needed
Best deal tracker for UGC creators?
One run- ChatGPT#2
- Perplexity#1
- Gemini#3
- AI Overviews#4
- ClaudeNot named
- GrokNot named
- AI ModeNot named
Sample data
In short
Tracking brand mentions in AI means asking AI engines a fixed set of buyer questions on a schedule and recording whether each answer names your brand, in what position, next to which competitors, and from which sources. You can do it by hand with a spreadsheet for a few prompts, or with Rankry, which automates it across up to nine AI engines and keeps every answer. The method is the same either way, and this guide covers both.
What to record
What should I track when monitoring my brand in AI?
A yes or no for “mentioned” is not enough. These six fields make each answer useful.
Named or not
Whether the answer mentions your brand at all. This feeds your visibility, the share of answers that name you.
Position
Where you appear among the brands named. First of five is a different outcome from fifth of five.
Competitors named
Which other brands appear, and in what order. This is how you see who AI favors for each question.
Sources cited
The pages and domains the answer links to, and whether any of them are yours.
How you are described
Whether the description is accurate, and whether it reads positive, neutral or negative. Save the answer text.
Context
The date, the engine and the exact prompt text. Without these you can’t compare one run with the next.
Manual method
How do I track brand mentions in AI by hand?
This method needs nothing beyond the AI products and a spreadsheet. The key is doing it the same way every time.
- 01
Write your prompt set
List 10 to 20 questions a buyer would ask, without your brand name. Include “best X for Y”, “alternatives to [competitor]” and use-case questions, and add a location if you sell locally.
- 02
Set up a sheet
One row per prompt and engine, with columns for date, named (yes or no), position, competitors named, sources, sentiment and a copy of the answer.
- 03
Use clean sessions
Log out or use a private window, and turn off memory and personalization where the product allows. Note the country you are asking from.
- 04
Ask each engine
Run every prompt in ChatGPT, Gemini, Claude, Perplexity and Grok. For Google, search the prompt and look for an AI Overview at the top, then ask it again in AI Mode.
- 05
Record, don’t interpret
Fill in the fields exactly as the answer shows them. If Google shows no AI Overview for a search, record that too: it counts as not named.
- 06
Compute your numbers
Visibility is answers naming you divided by all answers. Position is your average rank where named. Repeat both for each competitor.
- 07
Repeat on a fixed schedule
Daily gives you a fresh observation each day; 3-day and 7-day intervals use fewer requests. Keep prompts and engines unchanged so the trend means something.
Engine notes
How do I monitor my brand in ChatGPT, Gemini, Claude, Perplexity, Grok and Google?
Each surface has quirks that affect manual checks. Keep these in mind when you read the results.
| AI surface | Where to check | Watch out for |
|---|---|---|
| ChatGPT | chatgpt.com, logged out or with memory off | It may answer without searching, and then shows no sources. Answers vary between runs. |
| Gemini | gemini.google.com | It grounds answers in Google Search when needed; your Google account settings can shape results. |
| Claude | claude.ai, with web search on | Without web search, Claude answers from training alone and won’t reflect recent changes. |
| Perplexity | perplexity.ai | It searches almost every time and lists many sources; check whether you are named, not only cited. |
| Grok | grok.com or inside X | It can draw on posts from X as well as the web, so answers can shift with what people are discussing. |
| Google AI Overviews | A normal Google search for the prompt | An overview doesn’t appear for every search, and location and language change the results. |
| Google AI Mode | The AI Mode tab in Google Search | It runs several related searches per question and can cite different pages than AI Overviews. |
Manual vs tool
Should I track AI mentions manually or with a tool like Rankry?
Manual checks are a fine way to start. This comparison shows where they stop scaling.
| Aspect | Manual checking | Rankry |
|---|---|---|
| Cost | Free, apart from your time | 7-day free trial with no card, then a paid plan |
| Prompts | Whatever you can type; 10 to 20 is realistic | Written from your site at onboarding, plus your own or a CSV import, in categories |
| Engines | Any you have access to | ChatGPT, Gemini and Perplexity on every plan; Google AI Overviews and AI Mode on Pro or as Base add-ons; Claude, Grok, Meta Muse and Copilot as add-ons |
| How answers are collected | In the consumer app, shaped by your account, memory and location | Through each engine’s API with web search, one answer per prompt per run; Google surfaces from Google’s live results |
| Schedule | Whenever you remember | Re-runs automatically every day, or every 3 or 7 days if you choose; new prompts are answered right away |
| History | Your spreadsheet | Every run kept for your plan’s history window, 90 days by default, with the full answer for every new run |
| Competitors | You pick and count them yourself | Found automatically from your first report; add or remove brands and every figure recomputes instantly |
| Sources | Copied by hand from each answer | Every cited domain in the current report, with its type, a Yours, Gap or Mixed status and top URLs |
| Metrics | Whatever you calculate | Visibility, Position, Share of voice, Sentiment, Diversity and Rankry Score, per engine and over time |
| Next steps | Up to you | An Action plan, Tasks and Content Studio drafts built from the gaps |
Neither method shows “the” answer: AI answers vary by run, account and location. Consistency is what makes either method useful.
Why results differ
Why do AI answers about my brand change each time?
AI answers are generated, not looked up. The same prompt can produce a different list of brands from one run to the next, because models vary their wording, engines that search may retrieve different pages, and the products are updated often. In consumer apps, memory, past chats, location and account settings add more variation.
This is why a single check, yours or a colleague’s, can’t tell you how visible you are. Visibility is a rate measured over many prompts and repeated runs. If ChatGPT names you in three of ten answers this week and six of ten a month later, that change means more than any single answer.
It also explains why your own results can differ from a tracking tool’s. A tool asks in a consistent, neutral way, for example through an engine’s API without personal history, while you ask as yourself. Neither is wrong. Use one consistent method for trend data, and spot-check by hand when you want to see what a particular user might see.
Good practice
How do I keep AI mention tracking accurate?
Small habits make the difference between noise and a trend you can act on.
- Never include your brand name in the prompts you track
- Keep the prompt wording identical between runs
- Track the same competitors every time
- Record “not named” as carefully as “named”
- Save the full answer text, not just a yes or no
- Note the date and engine for every answer
- Look out for brand name variants and misspellings when reading answers
- Review trends monthly instead of reacting to single answers
- Add new prompts as a separate group so old trends stay comparable
With Rankry
How do I set up AI brand monitoring in Rankry?
Setup takes a few minutes, and the first report starts as soon as you confirm your prompts.
- 01
Enter your website
Rankry reads your site and fills in your brand, niche and location for you to review.
- 02
Review your prompts
Rankry suggests buyer prompts in categories, written from your site. Keep them, uncheck some, add your own or import a CSV.
- 03
Get your first report
Every prompt is asked on the engines your plan includes. Competitors are picked automatically from the brands the answers name.
- 04
Read the results
See Visibility, Position, Share of voice and Sentiment per engine, open any prompt to read its run history and full answers, and check the Sources page for gaps.
- 05
Act and re-measure
Use the Action plan to choose what to fix, then compare runs as reports re-run at your chosen interval: daily, every 3 days or every 7 days.
Guides
Keep reading
How do I track brand mentions in AI?
Write a fixed set of buyer prompts without your brand name, ask them in each AI engine on a regular schedule, and record whether you are named, your position, the competitors named and the sources cited. Use a spreadsheet for a handful of prompts, or a tool like Rankry to run and store it automatically.
How do I monitor my brand in ChatGPT?
Ask ChatGPT your buyer prompts in a logged-out or memory-off session, record whether you are named and which sources it cites, and repeat at a consistent interval with the same prompts. Rankry does this through OpenAI’s API with web search, daily by default or every 3 or 7 days if you choose, and keeps every answer.
Can I track brand mentions in AI for free?
Yes, by hand. The manual method in this guide costs nothing but time. It gets hard to keep up as you add prompts and engines, because each check multiplies: 20 prompts on 7 surfaces is 140 answers to read and record every round.
How often should I check AI mentions?
Rankry monitors daily by default, which helps you notice changes sooner. Choose every 3 or 7 days if you prefer a longer interval. AI answers vary between runs, so assess several observations together rather than reacting to one result. New prompts are answered immediately.
Which AI engines should I monitor?
The ones your buyers use. ChatGPT and Google’s AI Overviews and AI Mode are common starting points; Gemini, Perplexity, Claude and Grok matter depending on your market. Track several, since a brand is often named by some engines and not others.
Why does my brand show up for me but not in a tracking tool?
Your own sessions are shaped by your account, memory, chat history and location, and AI answers vary between runs anyway. A tool asks in a consistent, neutral way; Rankry, for example, asks through each engine’s API without personal history. Trust the trend over many answers, not one result.
What is the difference between a mention and a citation?
A mention is your brand named in the answer text. A citation is one of your pages linked as a source. You can be cited without being recommended, or recommended without a link. Track both, because they point to different fixes.
Does Rankry show the full AI answer?
Yes, for every new run. Rankry stores the full answer text alongside your position, the brands named, the sources and a sentiment chip. Runs from before full answers were stored keep the brands, the reasons the model gave and the sources.
Can I track my competitors’ mentions too?
Yes. Track the same prompts and record every brand named. In Rankry, competitors are found automatically from your first report; you can add other brands that appeared in your reports or remove ones you don’t compete with, and all figures recompute from stored answers without a re-run.
Can I track AI mentions for a local business?
Yes. Add the city or area to your prompts the way buyers do, for example “best dentist in Austin”. In Rankry, choosing Local at onboarding with your location makes prompts carry it and shows competitors with addresses on a map. Rankry does not track Google Maps or local-pack rankings.
Stop checking AI answers by hand
Rankry asks your buyer prompts on the AI engines your plan includes, keeps every answer and shows the trend.
7-day Base trial · No card needed