Key takeaways
- A free manual audit (10 to 15 buyer-style prompts, run several times, logged in a spreadsheet) gets you most of what a paid tracker would tell you at small scale.
- One prompt tells you almost nothing. Promptwatch Data found that 47.6% of ChatGPT responses cite 10 or more domains (Sep 28 to Oct 4, 2026), so you are usually sharing the answer with many others.
- Track more than yes/no mentions: position in the answer, how you are described, and which review sites and directories get cited next to you.
- Free analytics (GA4, Search Console, Bing Webmaster Tools) show whether AI visibility sends real visits.
- Free tiers and open-source trackers fill gaps, but once you run hundreds of prompt checks per month, a platform saves time.
Why local ChatGPT visibility is hard to measure
With Google, you check a rank. With ChatGPT there is no rank. The answer is generated, it changes between runs, and it may or may not name any business at all. A prompt like "best emergency plumber in Leeds" can give you a different shortlist on Monday than on Friday.
That noise is the main reason people reach for paid platforms. But the underlying job is simple: ask realistic questions, record what comes back, and look for patterns. You can do that for free. It just takes discipline.
One caveat before we start. I have not verified whether ChatGPT personalises answers by your physical location, and I would not assume it does. Treat it as something to test yourself: run the same prompt from your office, from a phone on mobile data, and with the city spelled out. If the answers differ, note it. If they do not, you have learned that city names in the prompt do the heavy lifting.
Method 1: the manual prompt audit
This is the core of any free approach. Backlinko's video on tracking your brand in ChatGPT and AI search for free walks through a version of it, built around mentions versus citations, an audit framework, data collection, and a 90-day action plan.

Step 1: write 10 to 15 real buyer prompts
Phrase them as questions, the way a customer would type them, not as keywords. For a local business, cover a spread:
- Organic, no brand named: "who is the best emergency plumber in Leeds?"
- Near-me style: "emergency plumber near me open now"
- Specific need: "how much does a boiler repair cost in Leeds?"
- Brand-specific: "is [Your Business] any good?"
- Comparison: "[Your Business] vs [Competitor]"
The mix matters. Promptwatch's unique-domains report shows brand-specific prompts behave differently from organic ones. For brand-specific prompts, five domains is the most common count and only 21.3% cite 10 or more. For organic prompts, 52.5% cite 10 or more. Branded prompts show you how you are described. Organic prompts show whether you get discovered.
Step 2: run each prompt several times
Run each prompt at least three to five times, ideally on different days and in fresh chats so earlier context does not leak in. Also try a logged-out session or a different browser profile. Promptwatch Data makes the same point: a single prompt tells you little, so measure across many.
If you want to see what ChatGPT is actually searching behind the scenes, the free Promptwatch GEO Inspector Chrome extension shows query fan-outs and which sources were fetched versus cited.
Step 3: log everything in a spreadsheet
A simple sheet is enough. Columns I would use:
| Column | What to record |
|---|---|
| Prompt | Exact wording, including city |
| Date and run number | Lets you spot variance |
| Mentioned? | Yes or no |
| Position | 1st, 2nd, 3rd in the list, or not listed |
| Description | How ChatGPT characterises you, copied verbatim |
| Sentiment | Positive, neutral, negative |
| Sources cited | Domains linked in the answer |
| Competitors named | Who appeared alongside you |
After 30 to 50 runs, calculate your mention rate (mentions divided by runs) per prompt type. That one number is your baseline.
Step 4: read the sources, not just the answer
The most useful column is the sources one. For local queries you will often see the same handful of domains: Google Business Profile pages, Yelp, Tripadvisor, industry directories, local news, Reddit threads. If a review site is cited for prompts about your brand, that site shapes how ChatGPT describes you. Fix your listing there before touching your own website.
Promptwatch Data shows what this looks like by industry. In its October 2026 travel citation report, booking and search sites earn more than half of ChatGPT citations. In its automotive report, marketplaces like Autotrader and Cars.com outrank every car brand. Local services work the same way: aggregators usually beat individual businesses, so being listed well on them is half the job.
Method 2: write prompts that match how ChatGPT searches
When ChatGPT browses the web, it turns your question into searches. Promptwatch's query fanout data measured about 2.15 searches per response in early December 2025, easing to about 1.84 by early March 2026, with average query length falling from roughly 117 characters to the high 80s. In April the measured figure was 1.0 per response at about 53 characters. The report's intro says a single prompt can trigger 3 to 8 or more searches, but the measured averages are much lower, so treat that as an upper range.
The practical takeaway: those searches look like short keyword queries, something like "emergency plumber Leeds". So test those short forms too, and make sure your Google Business Profile, page titles and headings put the service and the city up front.
Method 3: use your free analytics to see real visits
Mentions are nice. Visits and calls pay the bills. Free tools can show some of that.
Google Analytics 4
Look at Acquisition, then Traffic acquisition, and filter the source by chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and similar. You can also build a custom channel group with a regex on those sources so AI traffic gets its own line. Expect small numbers for a local business. A handful of visits that convert to calls can still matter.
I have not verified every referrer string, so check your own reports for what actually shows up and adjust the filter.
Search Console and Bing Webmaster Tools
Neither labels AI traffic cleanly, but both help indirectly. Search Console shows which queries and pages already perform, which are the pages most likely to be fetched by AI search. Bing matters more than people expect, since Microsoft Copilot runs on it. Verify your site there and submit your sitemap. Check Bing's own reporting for any AI-related views, as I could not confirm what is currently available.
Server logs
If you can access raw logs, search for user agents like GPTBot, OAI-SearchBot and ChatGPT-User. ChatGPT-User hits tend to show real people prompting ChatGPT in a way that fetched your page. Be careful: user agents can be spoofed, so a log line alone is not proof. Promptwatch publishes a directory of AI bots and crawlers so you can identify what you are seeing, and its AI crawler traffic data gives a benchmark for what normal looks like.
Also check that your robots.txt is not blocking these bots by accident. Plenty of sites are invisible to AI search simply because a security plugin turned it off.
Method 4: free tiers and free scans
Several tools offer a no-cost entry point. None replaces a full platform, but they speed up the manual work.
| Option | What is free | Catch |
|---|---|---|
| Promptwatch Explore plan | 1 project, 10 prompts, ChatGPT only | Small prompt limit |
| Promptwatch Free AI Visibility Report | No-login scan | One-off snapshot |
| Local Glyph | 30 free credits, no card (per its listing) | Paid packs run $45 to $500/mo |
| Local Dominator | 7-day free trial | Trial, not a free plan |
| GrowthOS | Free AI visibility report across ChatGPT, Claude, Gemini, Perplexity | Report, not ongoing tracking |
| Elmo | Open-source, self-hosted | You run the infrastructure |
| AI Monitor | Open-source brand mention tracker | Setup and API costs on your side |
Promptwatch has a free Explore tier and a no-login visibility scan, which is a reasonable place to see what automated tracking looks like before deciding whether to pay. Confirm current limits at promptwatch.com/pricing.

Elmo is a self-hosted option if you have someone technical around and want to avoid subscription fees.
AI Monitor is another open-source brand mention tracker for AI answers.

For local-specific work, Local Glyph and Local Dominator both advertise local AI visibility tracking. I have only seen their marketing pages, so check what the trial actually includes before relying on it. Local Dominator's own comparison lists Otterly Lite at $29/mo, which I did not verify elsewhere.
Otterly.AI is the budget paid option people mention most when the free route runs out.

Birdeye Search AI offers location-level tracking, which is relevant if you run several branches.

BrightLocal now has AI visibility alongside local rankings and listings.

Method 5: fix the free inputs that drive local mentions
Tracking is pointless if the inputs are weak. These cost nothing:
- Complete and consistent Google Business Profile, with the service and city in the description and categories.
- Matching name, address and phone number across Yelp, Bing Places, Apple Business Connect and industry directories.
- Real reviews on the platforms that get cited for your category, and replies to them.
- Service pages that put the service and the city in the title and H1.
- A short, factual FAQ on your site that answers the questions in your prompt list.
A note on Reddit and social. Promptwatch Data shows reddit.com's share of ChatGPT citations fell from about 4% to 0.5% on August 14, 2026 (report), and ChatGPT drew only 0.46% of citations from social platforms from Sep 28 to Oct 4, 2026 (report). Google AI Overviews, by contrast, got 12.2%. So do not bet your ChatGPT strategy on Reddit threads, though they may still matter for AI Overviews.
A simple monthly routine
Here is a schedule that takes about two hours a month for one location.
- Week 1: run your 10 to 15 prompts, three runs each, in ChatGPT. Log results.
- Week 2: repeat the same prompts in Google AI Overviews, Perplexity and Gemini, at least for your top five.
- Week 3: check GA4 for AI referrals and review your robots.txt and logs for AI crawler access.
- Week 4: compare mention rate to last month, note which review sites were cited, and pick one or two fixes.
Keep the prompt wording identical month to month. Changing it ruins the comparison.
Where the free approach breaks down
Being honest here: manual tracking has hard limits.
- Time. 15 prompts times 4 platforms times 3 runs is 180 checks a month, for one location.
- Consistency. Humans drift on wording and logging.
- Interface versus API. What you see in the ChatGPT app can differ from what an API returns, so scripts built on the API may not match what customers see.
- Scale. Ten locations or ten clients is not realistic by hand.
- Cause. A spreadsheet tells you what happened, not why. Crawler logs, citation analysis and page-level data answer that.
When you hit those limits, a platform starts to earn its price. Look for real-interface tracking, prompt volumes, citation sources, and some connection to traffic. Promptwatch goes further than monitoring, with crawler logs, visitor analytics and a content agent that acts on gaps, and its Essential plan starts at $95/mo with a 7-day free trial on monthly plans. If you are comparing options, the directories at bestgeosoftware.com and ai-rank-tools.com list many more.
Other trackers worth a look at the cheaper end are Peec AI and Rankscale.
Common mistakes
- Checking one prompt once and declaring victory or defeat.
- Only tracking branded prompts. Everyone finds themselves when they type their own name.
- Ignoring who else appears. Promptwatch Data recommends checking which competitors show up in the same responses as you.
- Treating a mention as the goal. Position and description matter, and a lukewarm description can cost you the click.
- Skipping the baseline. Without month-one numbers, you cannot show improvement later.
FAQ
Can I track ChatGPT visibility for free?
Yes, at small scale. Use a prompt list, repeated manual runs, a spreadsheet, and GA4 referral data. Free tiers like Promptwatch Explore add automation for a limited number of prompts.
How many prompts do I need?
Start with 10 to 15 that reflect real customer questions, including city-specific and near-me variants. Add more once the routine feels easy.
How often should I check?
Monthly is enough to see trends for most local businesses. Weekly is better if you are actively changing listings or content.
Does being cited mean I am recommended?
Not necessarily. Most responses cite many domains, so a citation may just mean your page was one of ten sources. Position in the answer and wording matter more.
Bottom line
Start with a spreadsheet, 15 prompts, and three runs each. Record mentions, position, description and cited sources. Pair that with GA4 referrals and a robots.txt check. Repeat monthly for a quarter, and by then you will know whether you need a paid platform or just better directory listings.


