1001 SEO Media
All posts
By 1001 SEO MediaGEOlocalBrightLocalAI searchprocess

BrightLocal AI Visibility Tracking: Local SEO and AI Search Features (2026)

How we use BrightLocal's Local AI Visibility module for clients already on BrightLocal, how we handle the 20-prompt cap and missing export, and when we switch.

When a local client comes to us with a BrightLocal account, we don't start by selling them a new tool. If they're on Track, Manage or Grow, they already have BrightLocal's Local AI Visibility module, included at no extra cost. Our first job is to use it properly. Our second is to be clear about when it stops being enough.

This post is how we do both. BrightLocal's details are from its pricing page, module page, llms.txt and help center, read on 20 October 2026.

What the module gives us

For each location, BrightLocal runs up to 20 prompts on ChatGPT, Google AI Mode and Google AI Overviews, with every query anchored to that location. It scores the location from 0 to 100 on how often it's mentioned, where it sits in the answer, how positively it's described and how strongly it's recommended. BrightLocal reads 75 and above as consistent recommendation and under 50 as mostly absent. Under the score sit mention rate, share of voice and average position by platform and prompt.

Two parts matter most to us. Competitors are detected from the answers themselves, so we see the businesses AI actually names near the client rather than the ones the client assumes it's up against. And the AI sources leaderboard shows which directories AI cites for the category and where the client is missing.

How we set it up

We don't accept the suggested prompts as they come. BrightLocal proposes them from the business category, which is a decent start, but categories are broad. A clinic that mostly sells one treatment needs prompts about that treatment. So we rewrite the list against the client's real services and the way their customers phrase things on calls and in reviews.

We split the 20 between branded and unbranded prompts, and we weight it heavily towards unbranded. Branded prompts tell you whether AI describes the business correctly. Unbranded prompts tell you whether a new customer would hear about it at all, and that's usually the question the client is paying us to answer. When we read results, we filter by prompt type first, so a strong branded score doesn't cover for a weak unbranded one.

We schedule weekly runs. Monthly is too slow to notice a change, and every on-demand refresh uses a manual report run, so we save those for checking a specific fix.

The sources leaderboard feeds straight into our directory work, the same process we describe in how we find which local directories AI cites. When a gap is worth fixing, BrightLocal can send it to Citation Builder, which is pay-as-you-go. We agree that budget with the client before ordering anything. We also tell them plainly that no AI engine documents that a new listing changes its answers. We order the listing, then watch the next few runs.

The 20-prompt cap in client reporting

Twenty prompts per location is a hard limit, and you can't buy more. You can swap prompts at any time, which is useful and also a trap. A prompt you remove stops collecting data, so its trend line ends. If we rotate prompts every month, the client's report becomes a list of snapshots with nothing to compare.

Our rule is to freeze the prompt set for a quarter. We keep a log of every prompt, when it went in and when it came out, so a change in the score can be traced to a change in the questions. For single-service clients the cap rarely bites. For clients with several services, or with towns they serve but don't have an address in, it bites quickly, because the module has nothing to anchor a prompt to without a location.

The no-export problem

BrightLocal's help center says export and white-label are on the roadmap, not live. In practice, that means the numbers stay inside BrightLocal. We have two options, and we don't love either. We can show the client the BrightLocal screen, or we can copy figures by hand into our own report.

Hand copying is where errors creep in, and our local reports have a fixed structure: one row per location per engine, citations, crawler errors, AI-referred visits. That structure is in how we report local AI visibility to clients. BrightLocal can fill the first section for three engines. The pages we read describe no AI crawler logs and no AI-referred traffic, so the rest of the report has to come from somewhere else anyway.

When we move the program to Promptwatch

We move a client's AI program to Promptwatch, the platform we run client programs on, when one of these comes up:

  • The client wants Perplexity, Gemini or Copilot in the report. BrightLocal's module doesn't track them.
  • The prompt set needs more than 20 questions, or covers towns without a location.
  • The report has to leave the tool, as a PDF, a CSV or a Looker Studio dashboard.
  • Someone asks whether AI answers brought in customers.
  • A score drops and we need to know why.

The last two are where Promptwatch does the work. Its crawler logs show which AI bots (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and others, verified by IP) fetched each location page and which visits failed. Its visitor analytics show AI-referred visits and conversions by page. Since ChatGPT stopped sending a referrer on 8 October 2026, we switch on the toggle that counts utm_source=chatgpt.com visits as ChatGPT, so those visits don't vanish into direct traffic.

For location, Promptwatch's agency plans include city monitors on every tier, and coordinate pins with a radius from 20 m to 5 km on Agency Scale. Brand clients who run their own account get city and coordinate monitors on Business. Scheduled PDF reports, CSV exports and the Looker Studio connector cover the reporting gap.

What stays in BrightLocal

We don't cancel BrightLocal when the AI program moves. It still runs the client's rank tracking, listings and, on Grow, review work. Citation Builder is still how we order listings. And we leave the AI module switched on as a free second reading on three engines, which costs nothing on the client's plan.

If you want to see what a fuller picture looks like before we talk, run one of your locations through Promptwatch's free AI Visibility Report. It's a no-login scan of ChatGPT, Gemini and Perplexity with a crawler access check, and it gives us a useful starting point for the first call.