How to Track Your ChatGPT Rankings Over Time

"How do I track my rankings on ChatGPT over time?" is one of the most common questions we get from teams who've only ever tracked Google. The honest answer is that ChatGPT doesn't have rankings in the Google sense at all — there's no fixed list of ten blue links to check your position on. But there is something you can track, consistently, that tells you whether your brand's standing in ChatGPT's answers is improving or slipping. Here's how to do it, starting with a spreadsheet you can build this afternoon.

Why ChatGPT "rankings" aren't like Google rankings

A Google ranking is a position on a page that (mostly) holds still between two people searching the same term at the same time. ChatGPT doesn't work that way, for three reasons that matter for how you measure it.

There's no fixed answer to check a position on. ChatGPT generates a new response each time, even for an identical prompt. Ask "what's the best project management tool for a 10-person agency" twice in a row and you can get two different lists, in a different order, with a different company named first. There's no URL you can bookmark and refresh.

Personalization and context shift the answer. A user's chat history, custom instructions, and account settings can all influence what gets recommended, the same way logged-in personalization has quietly shaped Google results for years — except ChatGPT's model has a lot more surface area to personalize against (a running conversation, memory settings, even the model version selected). Two people typing the identical prompt right now may not see the identical answer.

There's no rank-checking tool that reads the "real" result, because there isn't one canonical result. Classic rank trackers work by querying a search API and reading back position 1 through 10. ChatGPT has no equivalent API that returns "the ranking" — the only way to know what it's telling people is to actually ask it, repeatedly, and read what comes back.

Put together, this means single spot-checks are close to useless. Asking ChatGPT once, seeing your brand mentioned, and concluding "we're doing well on ChatGPT" tells you about one response, not your standing. The fix isn't a smarter single check — it's repeated sampling against a consistent prompt set, tracked over time, so the noise from any one answer averages out and the real trend shows through.

The manual way: a spreadsheet method

You don't need software to start. You need discipline and about 30 minutes a week. Here's the method.

1. Pick 20 buyer-intent prompts. These should be the questions your actual prospects would type into ChatGPT while shopping — not questions with your brand name in them. If you sell email marketing software, don't track "tell me about [YourBrand]." Track "best email marketing tool for a Shopify store," "email marketing software with good deliverability for cold outreach," "alternatives to Mailchimp for small teams." Aim for a mix: some broad category questions ("best X for Y"), some comparison questions ("X vs Y"), some problem-first questions ("how do I fix low email open rates"). Twenty is enough to see a pattern without turning this into a full-time job; fewer than ten and a single lucky or unlucky answer skews your whole read.

2. Ask each prompt, once a week, from a logged-out or fresh session. Consistency matters more than frequency here — same day, same 20 prompts, same conditions each time. Open a new chat for each prompt rather than continuing a thread, since ChatGPT can carry context from earlier in a conversation into later answers.

3. Log three things per prompt, every time:

• Named or not — a simple yes/no for whether your brand appears anywhere in the answer.
• Position in the list — if ChatGPT names multiple options, where do you land: first, middle, last, or "mentioned only if I ask for more options"?
• Who else is named — the other brands in the same answer, in order. This is what turns a mention log into a competitive one.

A basic spreadsheet with columns for prompt, week, named (Y/N), position, and competitors-named is enough. Add a short note field for anything the answer got wrong about you — outdated pricing, a discontinued feature, the wrong founding year — because those factual errors tend to repeat until you find and fix the source ChatGPT is pulling from.

4. Repeat weekly, without changing the prompt wording. The temptation is to tweak a prompt that "isn't giving good answers" — resist it. Changing the wording breaks the week-over-week comparison, which is the entire point. If a prompt turns out to be badly worded, retire it and start a new column, but don't retroactively edit history.

What to measure over time

Once you've got a few weeks of logged answers, four things are worth tracking as trends, not single data points:

Mention rate. Of your 20 tracked prompts, what percentage named your brand this week, and is that percentage moving up or down over the last month? This is the single most useful number, because it's the one least sensitive to any one weird answer.

Average position when mentioned. When you do show up, are you the first thing recommended, or the fourth item in a "some other options include" list at the end? Being named first-and-praised is worth far more than being buried, and a rising mention rate with a falling average position is a warning sign worth catching early.

Share of voice versus competitors. Across all your logged answers, how often is your brand named relative to the two or three competitors you compete with most? If a competitor's mentions are climbing while yours are flat, that's a relative decline even if your raw numbers look stable — share of voice is what actually matters to a prospect comparing options.

Citation sources. When ChatGPT's answer references where it got information — a review site, a comparison article, a forum thread — log those sources. Over a few months, you'll start to see which specific pages are shaping what ChatGPT says about your category, and whether they favor you or a competitor. That's actionable in a way "we got mentioned" isn't: you can pitch a correction to a review site, but you can't pitch a correction to "the model."

Pitfalls that quietly invalidate your data

A few mistakes are common enough to call out specifically, because each one can make weeks of tracking effort produce a misleading picture.

Checking once and calling it done. A single ChatGPT answer is one sample from a distribution, not a measurement. If you only check once a quarter, you have no way to tell whether a bad answer is a real decline or just the noise inherent to how ChatGPT generates responses.

Using branded prompts as your main signal. Asking "what does ChatGPT say about [YourBrand]" and being satisfied when it says something accurate tells you almost nothing about whether ChatGPT recommends you to someone who hasn't heard of you yet — which is the actual commercial question. Weight your prompt set toward unbranded, category-level questions.

Checking from a logged-in account, especially one with a long chat history about your own product. If you've spent months asking ChatGPT questions about your own company for research, testing, or content ideas, that history can bias what a "fresh" prompt returns in the same account. Use a logged-out session or a separate, clean account for tracking, and note which you used so a colleague can replicate the check the same way.

Treating one week's dip as a trend. Because ChatGPT's answers vary, a single bad week is expected noise. Look for a pattern across three or more consecutive checks before concluding something changed.

When to automate

The manual method above works, and plenty of small teams run it for months before it becomes a problem. It becomes a problem in a predictable way: the prompt list grows past 20 because one category isn't enough, someone wants to add a second or third AI platform because ChatGPT isn't the only place prospects are asking, and the 30 minutes a week quietly becomes half a day, done inconsistently, by whoever remembers.

If you're at that point, it's worth automating the repeated-prompt part specifically — running the same prompt set on the same schedule against the platforms you care about, and rolling up mention rate, position, share of voice, and citation sources the way described above, without someone manually copying answers into a spreadsheet.

If you just want a one-off baseline before committing to a weekly habit, our free AI visibility checker runs a quick check across several major AI platforms and shows you where you stand today — no spreadsheet required. When you're ready for the weekly version described above, done automatically, our ChatGPT rank tracker runs your prompt set on a schedule and tracks mention rate, position, share of voice, and citation sources over time, the same categories this post walked through — plus the other major AI platforms we track, and keep adding to, if ChatGPT visibility turns out to be only part of the picture.