Local Business Fact Check for AI Answers: Name, Address, Hours, Services

AI answers about a local business come from two places. Some come from what the model learned in training, which is frozen at a knowledge cutoff. Others come from pages the engine fetches at answer time: your site, your listings, and directories that mention you.

Each path fails in its own way:

• Training-era answers can repeat hours or an address you changed long ago. Nothing you publish this quarter reaches them.
• Retrieved answers can quote a stale directory instead of your site, or pick one of two addresses that you published yourself.
• Retrieval is per query. As of October 2026, engines decide case by case whether to search the web, so the same question can get a different answer tomorrow.

Knowing whether an error is a hallucination or an outdated answer changes the fix. What sits in your control is narrow and concrete: state each fact once, the same way everywhere, on pages crawlers can read. For the wider picture, see how local businesses show up in AI answers.

Say you run Riverside Family Dental, a made-up clinic with two locations in one city. Nothing about it is real. It exists to show the kinds of drift a normal business builds up.

• In March the main location moved from Suite 200 to Suite 310. The contact page says 310. The footer and two directories still say 200.
• Saturday hours ended in the summer. The location page is updated, but the hours live in a booking widget that loads with JavaScript.
• The clinic stopped offering orthodontics. An old services page still lists it and is still indexed.
• The homepage says it serves "the metro area". One listing names two specific suburbs. Another names a county.
• The second location's phone number appears in three formats, and one is a retired tracking number.

Every item is a fact an AI engine can repeat wrongly, and every item is fixable by the clinic. None of them needs a new tactic. They need a single source of truth and a crawl-friendly page that states it.

Run these steps once per location. A spreadsheet is enough.

1. List the facts. One row each for public name, street address with suite, phone, hours for every day, holiday hours, current services, dropped services, and service area.
2. Write the correct value. Use the exact wording you want quoted. Name service areas as specific places, not "nearby".
3. Map where each fact appears. Homepage, contact page, location pages, footer, booking widget, any structured data block, your map and business listings, review sites, directories and social profiles.
4. Mark every mismatch. A different suite number, a 24-hour versus 12-hour time format that changes meaning, or a service you no longer offer all count.
5. Fix in order. Your own site first, then listings you control, then requests to third-party sites that you don't control.
6. Date the sheet. Record when each fact last changed. You'll need it to judge whether a wrong AI answer is stale or invented.

The free check for contradicting facts on your own site speeds up step four for the pages you own.

Work through this for each location. Tick an item only when every surface matches the fact sheet.

Name

• The same public name on the site, every listing and every profile
• No old trading names left on pages that are still indexed
• Location names that a customer would recognise ("Riverside Family Dental, North")

Address

• Street, suite and postcode identical everywhere
• Old addresses removed or clearly marked as former, not left as a second current address
• Each location on its own page with its own address in plain text

Phone

• One number per location, in one format
• Retired or tracking numbers removed from listings

Hours

• Every day stated, including closed days
• Holiday and seasonal hours published before they start, then removed after
• Hours present in the page HTML, not only inside a booking or map widget

Services

• A current services list per location
• Dropped services removed, or the page says plainly that you no longer offer them
• Specific facts stated in one sentence where it matters, such as "We see emergency patients the same day, Monday to Friday"

Service area

• Named places, the same list on the site and every listing
• No conflict between "metro area" on one page and a short suburb list on another

A plain, specific sentence on a fetchable page gives an engine something it can quote. That's the most direct thing on-page content can do.

Consistent facts still fail if crawlers can't fetch them. These are the access checks with the strongest evidence behind them.

• JavaScript. Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. Hours inside a script-loaded widget may be invisible to them. Put the same hours in the server-rendered HTML.
• robots.txt, per bot. Rules apply bot by bot. OpenAI documents that GPTBot governs training, while OAI-SearchBot controls whether you appear in ChatGPT search, and that each setting is independent. A blanket AI block can remove you from search features you wanted.
• Firewall and CDN. A firewall can return 403 to a crawler that robots.txt allows. Nothing in your dashboard flags it.
• Snippet and index controls. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers. Check that a template hasn't applied them to your location pages.
• Index eligibility. Google states a page must be indexed and eligible to show with a snippet to appear as a supporting link in AI Overviews or AI Mode.

A minimal robots.txt that lets the main search crawlers in looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

To confirm the firewall half, the AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.

Now test what the engines actually say. Ask the questions a customer would ask, logged out where you can, and record every answer.

1. Write five to ten questions: "Is Riverside Family Dental open Saturday?", "Where is the north location?", "Do they do orthodontics?"
2. Ask each engine separately: ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity, Claude and Copilot.
3. Log the answer, whether it cited a source, and which source.
4. Compare each fact against your dated fact sheet.

The free checker on this site runs up to three buyer questions as a snapshot if you want a quick start.

Engines can disagree with each other. In Otterly's 2026 experiment, Google removed two test sites from its index and Google's AI surfaces almost stopped citing them, while Bing kept them and Microsoft Copilot kept citing them. Check each engine on its own. For engine-specific fixes, see checking Google's AI Overview for accuracy and fixing wrong facts in Microsoft Copilot.

Local facts change more often than most brand facts. Set a rhythm that matches that.

• The week any fact changes. A move, new hours, a dropped service: update the fact sheet, the site and every listing, then re-run your questions a few weeks later.
• Before seasonal or holiday hours. Publish them early and remove them after. Businesses where hours shift often, like restaurants managing menus and hours, need this most.
• Monthly. A quick spot check: three questions per engine on hours and address.
• Quarterly. The full checklist, the crawler checks and a fresh pass through every listing.

Expect some errors to outlast your fixes. Retrieved answers can update once engines re-fetch the corrected pages. Training-era answers change only when a model is retrained, and no one outside the vendor controls that timing. Your job is to make sure that every page an engine might fetch tells the same, correct story.