How to Test Whether AI Recommends Your Business

Most companies are guessing about AI visibility.

They ask one question in ChatGPT, see one answer, and call it research.

That is not research.

That is a screenshot.

If you want to know whether AI recommends your business, you need a repeatable test. Not a perfect test. Not a magical test. A practical one.

The goal is not to prove that one AI tool always behaves the same way. It will not.

The goal is to see patterns:

• Are you named?
• Are competitors named?
• Are directories named instead?
• Are you cited as a source?
• Are you absent from the buyer questions that should belong to you?
• Do answer engines understand what you do?
• Do they connect you to your service area?
• Do they trust third-party sources more than your own site?

That is where the signal lives.

Short answer

To test whether AI recommends your business, run a controlled set of buyer-style prompts across multiple answer surfaces, document whether your business is named, cited, mentioned, skipped, or replaced by a directory, then compare the results against your website, Google Business Profile, reviews, structured data, and citation ecosystem.

The test should be repeated over time.

One prompt is noise.

A structured prompt set is intelligence.

What you are actually testing

You are not only testing "whether ChatGPT likes you."

That is the wrong frame.

You are testing whether the web gives AI systems enough clear, verifiable information to include your business in high-intent answers.

That includes:

• Entity clarity
• Service clarity
• Local relevance
• Review specificity
• Third-party corroboration
• Website usefulness
• Structured data
• Directory consistency
• Question coverage
• Conversion path clarity

AI visibility is not a single ranking.

It is a pattern across search surfaces.

The difference between mention, citation, and recommendation

Before testing, define the outcomes.

Mention

The AI system names your company but does not link or clearly endorse it.

Example:

Companies in the area include ABC Plumbing, Smith Plumbing, and Metro Rooter.

A mention is useful, but weaker than a citation or recommendation.

Citation

The AI system links to your website, profile, article, directory listing, or another source that includes your company.

Example:

According to ABC Plumbing's emergency service page...

A citation shows the system found a source it considered usable.

Recommendation

The system includes your company as an answer to a "who should I call" or "best option" query.

Example:

For emergency plumbing in Fort Worth, consider ABC Plumbing because they list 24/7 service, have strong recent reviews, and serve the area.

A recommendation is the highest-value outcome.

Directory replacement

The system does not name your business but names a directory or aggregator instead.

Example:

Check Angi, Yelp, or HomeAdvisor for emergency plumbers near you.

This is a warning sign.

It means the system may find directories easier to use than your own digital presence.

Skip

The system names competitors but not you.

This is the most important failure state.

If buyers ask a question that should belong to your business and AI names three competitors, you have a visibility problem.

Step 1: Define the buyer decision

Do not test random prompts.

Start with the decision you want to own.

For a roofer:

Homeowner has storm damage and needs to know who to call.

For a plumber:

Homeowner has water on the floor and needs emergency help.

For HVAC:

Homeowner's AC dies in July and they need same-day service.

For an electrician:

Homeowner has a safety concern and needs a qualified professional.

For a premium technical brand:

Buyer is comparing products or trying to choose the right model for a use case.

The prompt set should map to the buyer's decision, not your internal service list.

Step 2: Build a prompt set

A strong prompt set includes multiple types of buyer questions.

Direct recommendation prompts

These ask for a company or provider.

Examples:

• Who is the best roofer near me for storm damage?
• Which emergency plumber should I call in Fort Worth?
• What HVAC company can fix my AC today in Plano?
• Who installs EV chargers near me?
• Which foundation repair company is trusted in North Texas?

Problem-first prompts

These ask what to do.

Examples:

• What should I do when a pipe bursts?
• What should I do after hail damages my roof?
• Why is my AC blowing warm air?
• What should I do if my electrical panel smells like burning?
• What do stair-step cracks mean in a brick wall?

Comparison prompts

These ask the system to weigh options.

Examples:

• Roof repair vs. roof replacement after hail damage
• Plumber vs. restoration company after a pipe burst
• AC repair vs. AC replacement
• Panel upgrade vs. electrical repair
• Foundation repair methods for North Texas soil

Trust prompts

These ask how to choose well.

Examples:

• How do I choose a trustworthy roofer?
• What should I ask before hiring an emergency plumber?
• How do I compare HVAC companies?
• What credentials should an electrician have?
• How do I avoid foundation repair scams?

Local prompts

These tie the query to geography.

Examples:

• Emergency plumber in Fort Worth open now
• Storm damage roofer in Dallas
• HVAC repair in Plano today
• Tree removal after storm in Arlington
• Foundation repair in North Texas

Do not test only branded prompts.

If you ask, "Tell me about my company," you are not testing buyer discovery.

You are testing brand recall.

The money is in non-branded prompts.

Step 3: Choose the surfaces to test

Test across multiple answer surfaces.

At minimum:

• Google Search with AI features where available
• Google Maps
• Perplexity
• ChatGPT with web access if available
• Gemini
• Bing/Copilot if relevant
• Voice assistant queries if relevant

Each surface behaves differently.

Google may lean on its index and Maps ecosystem.

Perplexity may emphasize citations.

ChatGPT may synthesize across sources differently depending on access and context.

Gemini may interact with Google's broader ecosystem.

Voice assistants may compress results even further.

That variation is the reason testing one tool is not enough.

Step 4: Control the test conditions

AI outputs vary.

That does not make testing useless.

It means the test needs guardrails.

Record:

• Date
• Tool
• Model or interface if visible
• Prompt
• Location setting if available
• Browser/session notes
• Whether you were logged in
• Whether the tool used web search
• The exact answer
• Sources/citations shown
• Companies named
• Directories named
• Whether your company appeared
• Screenshots

Run the same prompt set more than once over time.

Do not overreact to one output.

Look for repeated patterns.

Step 5: Score the result

A simple score is enough to start.

Use this:

0 — Not visible

Your company is not mentioned, cited, or recommended.

1 — Weak mention

Your company is mentioned, but not linked, cited, or recommended.

2 — Cited source

Your company, website, or profile is cited as a source.

3 — Included in shortlist

Your company appears in a list of recommended or relevant providers.

4 — Strong recommendation

Your company is named with supporting reasons tied to service, location, reviews, proof, or relevance.

Also track whether a directory or competitor appeared instead.

That is often more useful than the score.

Step 6: Compare AI output to Google Maps

For local businesses, compare answer results to Maps results.

Ask:

• Does the AI answer match the Map pack?
• Does the AI name companies that rank in Maps?
• Does it cite directories instead of Maps?
• Does it name companies with more reviews?
• Does it name companies with more complete profiles?
• Does it ignore companies that appear strong in Maps?

This comparison reveals whether your local entity signals are aligned.

If you rank in Maps but do not appear in AI answers, that is a different problem than not appearing anywhere.

Step 7: Diagnose why you were skipped

Do not stop at "we did not appear."

Ask why.

Look for gaps.

Entity gaps

• Inconsistent business name
• Old phone numbers
• Conflicting directories
• Weak or missing Google Business Profile data

Service gaps

• No dedicated service page
• Vague service language
• Missing emergency/same-day availability signals
• No page for the exact prompt being tested

Proof gaps

• Generic reviews
• Few recent reviews
• No service-specific testimonials
• Missing project proof
• No credentials visible

Content gaps

• No answer-ready pages
• No question clusters
• No comparison content
• No local pages
• No videos or transcripts

Technical gaps

• Missing schema
• Bad sitemap/canonicals
• Thin pages
• Poor internal linking
• Slow mobile experience
• Non-indexable pages

Source gaps

• Few third-party mentions
• Missing directories
• No associations
• No YouTube/LinkedIn/industry presence
• Aggregators are clearer than your own website

The output is the symptom.

The signals explain the cause.

Step 8: Track competitors and directories

Every test should capture who appears instead.

Do not only track yourself.

Track:

• Competitors named
• Directories named
• Sources cited
• Types of pages cited
• Review counts
• Service pages
• Google Business Profile strength
• Content depth
• Structured data
• Local relevance

The competitors that appear repeatedly are not random.

They may be sending clearer signals.

Or a directory may be controlling the answer layer in your category.

Both matter.

Step 9: Repeat monthly

AI visibility is not static.

Models change.

Search interfaces change.

Competitors publish.

Reviews accumulate.

Directories update.

Google Business Profile changes.

Your own site changes.

A one-time test is a snapshot.

A monthly test becomes a trend.

For most local businesses, a monthly prompt test is enough to start.

For high-value categories or aggressive markets, test more often.

A practical prompt-testing sheet

Track these columns:

• Date
• Market
• Trade/category
• Prompt
• Tool
• Location setting
• Your company named?
• Your company cited?
• Your company recommended?
• Competitors named
• Directories named
• Sources cited
• Answer summary
• Visibility score
• Likely gap
• Recommended fix
• Screenshot link

This turns AI visibility into something you can manage.

Example: emergency plumber test

Prompt:

Who should I call if a pipe bursts in Fort Worth tonight?

Possible findings:

• Your company does not appear.
• Two competitors appear.
• Yelp appears.
• One competitor has an emergency plumbing page.
• The other has reviews mentioning "same day" and "burst pipe."
• Your Google Business Profile lists plumbing but not emergency service.
• Your site has no burst pipe page.
• Your reviews are positive but generic.

The fix is not "do more SEO."

The fix is specific:

• Build an emergency plumbing page
• Build a burst pipe question cluster
• Update Google Business Profile services
• Improve review request language
• Add FAQ/schema where appropriate
• Clean local citations
• Add internal links
• Retest the prompt set

That is how testing becomes strategy.

What not to do

Do not manipulate prompts to get the answer you want.

Do not run only branded prompts.

Do not cherry-pick one good answer and call it proof.

Do not promise that optimization will guarantee recommendations.

Do not use fake best-of pages or deceptive content.

Do not spam Reddit, Quora, or directories.

Do not turn AI visibility testing into theater.

The goal is not to manufacture screenshots.

The goal is to find the truth.

Final answer

To test whether AI recommends your business, you need a repeatable prompt-testing system across multiple answer surfaces.

Test real buyer questions.

Track mentions, citations, recommendations, skips, and directory replacements.

Compare results to Maps, reviews, service pages, schema, citations, and content gaps.

Then improve the signals and test again.

That is how AI visibility becomes measurable.

Want us to run the test for you?

Book a 6Signal Visibility Audit.

We'll test your company across AI answers, Maps, search, directories, and voice to show where you appear, where you get skipped, and what to fix first.

Sources and further reading

• Google Search Central: AI features and your website
• Google Search Central: Google Search Essentials
• Google Business Profile: Tips to improve your local ranking
• Google Search Central: Introduction to structured data markup
• Schema.org: Organization, LocalBusiness, Service, FAQPage, BreadcrumbList