How to conduct local GEO research

Ask 10 local business owners how they’re doing in AI search, and nine will point to their Google Business Profile. That’s the wrong place to look.

ChatGPT recommended only 1.2% of the nearly 350,000 business locations analyzed in SOCi’s 2026 Local Visibility Index. Compare that to the 35.9% view rate those same brands get on Google’s local 3-pack, a nearly 30-fold gap. Gemini graced 11% of the places. Confusion, 7.4%.

Business profile information across the web was about 68% accurate for ChatGPT and Perplexity, compared to 100% for Gemini, which relies entirely on Google Maps data.

A business can dominate the map pack and still disappear when someone asks ChatGPT for a recommendation. Many local businesses have never looked at what AI has to say about them, so they invest in content and citations without knowing where any of the impressions actually count.

GEO’s location-based assessment fixes that. It gives you a repeatable way to measure how AI describes, recommends, or ignores a business before you spend a dollar trying to improve it. Here’s how to run one.

Why the foundation comes first

Think of it as stepping on the scale before starting to eat. If you don’t know your number one, you won’t know if what you’re doing is working. The foundation gives you numbers to track over time: share of voice, citation rate, and accuracy.

It also answers the big question: Can AI crawl, understand, and trust this site? The answer changes everything you do next. That’s why you need to uncover relevance issues before thinking about a content strategy.

AI also measures signals differently than traditional local search. Traditional local search relies heavily on proximity. Whoever is closest usually wins. AI doesn’t play that game. It prioritizes data trust, authority, and consistency across the web.

Third-party verification and accurate, consistent business information are more important than how far you search. AI often relies on the same business data as traditional local search, which is weighted differently. That’s why map pack levels don’t tell you anything about AI visibility.

Step 1: Gather your research input

Before using a single prompt, edit. Open the spreadsheet and close the four categories of questions because each one reveals a different type of weakness:

• Discovery: “the best [service] next to me” or “up [service] in the middle [city]”
• Comparison: “[Brand] vs. [Competitor] in the middle [city]”
• Trust: “[Brand] updates” or “of course [Brand] reliable?”
• Transportation: hours, address, parking and phone number

Use each question on the AI ​​platforms your customers actually use: ChatGPT, Perplexity, Gemini, and Google AI Overviews. Each platform pulls from different sources and phrases respond differently, so appearing on one doesn’t guarantee you’ll appear on another.

A few exceptions can silently harm your data if you don’t control them. The AI ​​answers change based on who is asking, so check in the defined area and note the exact city or ZIP code.

Run a clean, logged-out session next to the login to minimize personalization noise. Date stamp every run. These models are updated regularly, and last month’s screenshot doesn’t tell you much other than the date attached.

Dig deep: The new playbook for local AI search optimization

Step 2: Run the commands and record the results

For every command on every platform, capture five things:

• Speak: Does AI mean business?
• Specify the order: First, middle, last, or none?
• Emotions and framework: Good, neutral, or bad?
• True accuracy: Are the hours, services, and prices correct?
• Sources cited: What URLs and directories does the answer rely on?

Those five data points will tell you more than most agencies ever rate a client.

• Set up your spreadsheet with columns for:
  • Immediately.
  • The platform.
  • Say it.
  • Position.
  • Accuracy score.
  • Emotions.
  • Quotation calculation.
  • Top sources.
• Then calculate two summary metrics:
  • Visibility percentage (how often the business appears).
  • Percentage of accuracy (how many times the facts are correct).

If you’d rather not build this from scratch, here’s a free template that includes a response log, competitor tracker, scorecard, and gap tracker. The link opens your editable copy.

While you’re at it, log competitors, too. Note who else appeared in each answer, where they rated them, and what sources supported them. That tells you who is winning the category in the eyes of the AI, and usually why.

Step 3: Find the spaces

Each gap you receive falls into one of three buckets:

• Invisible: Business just doesn’t appear in the related question. This is the most common failure mode for local businesses that are just starting to explore AI visibility. It often traces back to blocked searches, a lack of available content, or a few third-party chats.
• Incorrect: Business is emerging, but the details are wrong. Old address. The hours are not the same as reality. The service was discontinued two years ago. This is not just anger. AI treats inconsistency as a signal of trust, so businesses with unreliable information are more likely to be blocked or excluded from responses. This often traces back to outdated site information or inconsistent NAP data (name, address, phone) across directories.
• Incorrect framing: Business is mentioned, but it is buried under competitors or described as a weak option. This usually comes from a smaller review profile or weaker authority signals than competitors who win that query.

Once you know which bucket the gap falls into, you can prioritize the right fixes.

Dig deeper: How to create FAQs that power AI-driven local search

Step 4: Adjust accordingly

Consistency is more important than people expect. Skip ahead, and you’ll waste the job.

Fitness comes first

• Can AI crawlers even access the site? Check robots.txt and any Cloudflare settings that may block bots automatically. Cloudflare announced that it would automatically block AI crawlers from sites on its network, and many site owners didn’t notice that the setting had changed.
• Clean up NAP consistency so that your business name, address, and phone number are the same everywhere on the Internet.
• Add and validate structured data, including LocalBusiness, Organization, FAQ, and service schema markups.

Dig deeper: The 90-day SEO playbook for AI-driven search visibility

Signs of secondary dependence

• Build a strong review profile and healthy average ratings across Google Business Profile, Yelp, and industry-specific sites.
• Respond to reviews and questions. AI sees engagement, not just star ratings.
• Identify platform relevance, with the same story displayed across your website, directories, social profiles, and media coverage. Consistency shows trustworthiness.

The ultimate compatibility

Only now does content work make sense.

• Build real depth of location with city pages that feature real location information, service pages with real examples, and clear travel information.
• Skip the cookie cutter pages that just switch to another city name.

The logic behind the order is simple. If searches are blocked or your NAP data is inconsistent, the AI ​​may never see new content. Developing relevance before eligibility is like repainting a house that no one can find.

Step 5: Do the research again

One test is a snapshot. Real progress comes from iterating on schedule because AI models are constantly being updated, and what’s true today may not hold up next quarter.

A quarter is a reasonable situation for many local businesses. That’s usually enough to catch model updates and measure whether your fixes are working without turning the process into a full-time job.

Don’t focus on clicks as the primary success metric. That trend continues from traditional SEO, but it doesn’t show a clean map of how AI responses work.

Watch for name search suggestions, more calls, and more referral requests from local profiles instead. Those signals show whether AI recommendations are driving real business, even if there are no clicks to be attributed.

Within the research itself, track mention rate, positioning, factual error rate, and citation count over time. If the mention rate goes up but the position remains buried below our competitors, that’s a trust issue, not a fit issue, and it should shape your priorities for the next quarter.

Compare each study with the last. If the AI ​​suddenly chooses different sources or phrases that respond differently, that’s a model drift that needs to be understood, not just one mistake you can ignore.

Keep track of your competitor’s voice shares, too. If a competitor is slowly moving up through the same order, it’s worth investigating before it takes off. Sometimes it’s a real power gap. Sometimes, they just start answering updates while everyone else is waiting.

Dig deeper: Local SEO sprints: A 90-day plan for service businesses in 2026

Start here, not with content

Basic GEO testing is not complicated. Benchmark where things stand, fix relevance and trust issues before touching content, organize your information so AI has a reason to cite the business, and repeat tests on a regular schedule.

Another guess. In local search, guesswork appears as missed calls and customers who choose a competitor without knowing why. If no one has looked at what AI has to say about business lately, this is a place to start.