GEO Diagnostic
The 20-Prompt AI Visibility Test: Find Out If ChatGPT and Perplexity Understand Your Business
One prompt is a curiosity. Twenty consistent prompts are a diagnostic. If AI answer engines cannot explain who you help, what problem you solve, where you operate, and why a buyer should trust you, your website probably has a clarity problem before it has an AI problem.
Direct answer: what is a 20-prompt AI visibility test?
The 20-prompt AI visibility test is a fixed monthly check that asks ChatGPT, Perplexity, Gemini, Copilot, and Google AI experiences the same buyer questions. Record whether your business appears, is cited, is described correctly, and sends people to useful pages. The point is diagnosis, not vanity scoring.
This test is a simple way to find out whether AI-assisted search systems understand your business in the same way a good buyer should. It will not prove a perfect ranking. It will not guarantee that your company appears every time someone asks a related question. It will show the gaps that matter: unclear services, weak proof, missing comparison content, thin local signals, and pages that are not citeable enough.
If you are new to this work, start with the broader Generative Engine Optimization guide. If you already know visibility is weak and want a practical improvement map, Run a GEO Readiness Audit.
What the AI visibility test measures
An AI visibility test measures how answer engines handle buyer-like questions about your market. The useful signals are not just whether your brand name appears. You also want to know whether the answer system understands your category, your best-fit customer, your service area, your differentiators, your proof, and the next step a buyer should take.
For a small business owner, this matters because AI search is not only a new traffic source. It is a new interpretation layer. A buyer may ask ChatGPT, Perplexity, Gemini, Copilot, or Google AI features to compare options, explain what to look for, shortlist providers, or summarize a service category. If your public information is vague, the answer may skip you or describe you in a way that does not help.
The test belongs next to your monthly reporting. The previous article on how to measure GEO performance explains the broader scorecard: rankings, citations, mentions, referrals, prompt-test accuracy, and business outcomes. The 20-prompt test is the practical prompt-test part of that scorecard.
Why this matters for SMB owners
Most SMB websites were built for human skimming and Google search. That is still important. But answer engines also need clear, extractable information: what you do, who it is for, where you work, what the process looks like, what proof exists, and which page supports each claim.
Google says its AI features use the same broad SEO foundations: crawlable pages, helpful content, internal links, high-quality media, textual information, and structured data that matches visible content. OpenAI's publisher guidance also makes crawler access and referral tracking practical concerns for site owners who want to appear in ChatGPT search experiences. Bing's AI Performance report gives another useful signal by showing pages cited in AI-generated answers and the grounding queries associated with them.
That is the business reason to test. The test does not replace SEO, analytics, or sales conversations. It gives you a monthly view of how your public information is being interpreted. For an owner, that can be the difference between guessing and knowing what to fix next.
Before you run prompts, check the basics
Do not start by arguing with an AI answer. Start by checking whether the site gives the answer engine enough clean material to work with.
Pre-test readiness checklist
- Confirm your priority pages are publicly accessible and not blocked by robots.txt or noindex rules.
- Make sure OAI-SearchBot is not blocked if ChatGPT search visibility is a goal.
- Pick five to ten priority URLs: homepage, main service pages, local pages, case examples, comparison pages, and your audit or booking page.
- Check whether each priority page has a clear answer block, service explanation, buyer example, proof section, FAQ, and next step.
- Set up a simple spreadsheet with columns for prompt, tool, brand mention, citation, competitor names, accuracy, landing page, and next action.
- Decide what "good enough" means before you test. Otherwise you will move the goalposts after every answer.
If you want a deeper starting checklist, use the AI Search Visibility Audit as the companion article. The prompt test is stronger when the website is already crawlable, structured, and commercially clear.
The 20 prompts to test your AI search visibility
Use the prompts below as templates. Replace the bracketed details with your industry, city, service area, buyer type, and service category. Run them in the same tools each month. For most SMBs, I would test ChatGPT, Perplexity, Gemini, Copilot, and Google AI experiences where available.
1. Category discovery prompts
- What should a [buyer type] look for when hiring a [service category] in [city or region]?
- Who are good [service category] providers for [buyer type] near [city or region]?
- What are the best options for [specific problem] if we are a [business type] with [constraint]?
- Which companies are known for helping [buyer type] with [specific outcome]?
2. Problem and solution prompts
- How can a [business type] solve [pain point] without adding another full-time role?
- What is the practical process for improving [problem] in a small business?
- What questions should I ask before buying [service or solution]?
- What mistakes do small businesses make when trying to fix [problem]?
3. Comparison and trust prompts
- Compare [your category] options for a [buyer type] that needs [outcome].
- What makes one [service category] provider more credible than another?
- What proof should I expect before hiring a [service category] consultant or agency?
- What are the tradeoffs between hiring a consultant, agency, or internal employee for [problem]?
4. Brand-specific prompts
- What does [your business name] do?
- Who is [your business name] best suited for?
- What services does [your business name] offer for [buyer type]?
- Can you summarize [your business name] and cite the best source?
5. Commercial-intent prompts
- I need help with [problem]. Which pages should I read before booking a call?
- What should a [buyer type] prepare before asking for a [service category] audit?
- Which provider looks most relevant for [specific use case], and why?
- If I only have one month to improve [problem], what should I fix first?
How to score the test without fooling yourself
Keep the scoring simple. A complicated score can make weak evidence look precise. You need enough structure to see patterns, not a fake scientific grade.
| Signal | Score it this way | What it usually means | Next action |
|---|---|---|---|
| Brand mention | 0 = absent, 1 = mentioned, 2 = mentioned naturally with context | The answer engine recognizes your business for the topic. | Improve service-page clarity and external proof if absent. |
| Citation | 0 = no source, 1 = weak source, 2 = priority URL cited | Your page is usable as supporting evidence. | Strengthen answer blocks, examples, FAQ, and source-ready sections. |
| Accuracy | 0 = wrong, 1 = vague, 2 = accurate and buyer-useful | Your public language is clear enough to interpret. | Rewrite vague positioning and service descriptions. |
| Competitor pattern | Record names and cited pages | Shows which pages answer engines find easier to use. | Compare structure, proof, specificity, and recency. |
| Commercial next step | 0 = none, 1 = generic, 2 = useful next page or action | The answer can move a buyer toward a decision. | Add a clear CTA, audit page, comparison page, or service page. |
Do not overreact to one bad response. AI answers vary. Run the same test monthly, record the same fields, and look for repeated weaknesses. If the same service is misunderstood across tools, that is a website clarity issue. If competitors are cited repeatedly and you are not, study the cited pages. If your brand appears but the description is wrong, fix your own public language first.
This is also where the earlier article on AI answer engine citations and brand mentions becomes practical. A mention can be useful, but a citation to the right page is stronger because it gives the buyer a path back to your source.
How to interpret the most common gaps
Your business is absent for category prompts
This usually means the answer engine does not see enough clear, trusted, public information connecting your business to the category. Check the basics: page titles, service headings, local signals, industry examples, case studies, public profiles, and internal links. A vague homepage is rarely enough.
Your business appears, but the description is generic
This is common. The answer says something like "provides business solutions" or "helps companies with digital transformation." That sounds polite, but it does not help a buyer. Rewrite your service pages with concrete workflows, buyer situations, industries served, constraints, and outcomes. The article on making your website more citeable by AI search engines explains how to make those sections easier to reuse.
Competitors are cited and you are not
Do not treat this as an insult. Treat it as research. Open the cited competitor pages and look at structure. Do they have clearer headings? Better examples? More precise service descriptions? Better FAQ answers? More recent updates? Stronger external proof? Your job is not to copy them. Your job is to make your own best information easier to find, understand, and cite.
AI referrals exist, but leads do not improve
Then the traffic may be landing on weak pages, or the next step may be unclear. Use Google Analytics traffic-source dimensions to review where visitors came from, which landing pages they used, and whether they completed useful events. For GEO work, the question is not just "Did AI send traffic?" It is "Did the right visitor land on a page that helped them take the next step?"
A practical US SMB example
Imagine a 12-person dental marketing agency in Ohio. The owner wants to appear when a dental practice asks an AI tool: "Who can help us get more implant patients without wasting money on broad ads?"
The agency runs the 20-prompt test. For broad category prompts, it appears once in Perplexity but not in ChatGPT or Copilot. For problem prompts, competitors appear because they have specific pages about implant patient acquisition, call tracking, landing-page conversion, insurance-sensitive copy, and lead quality. The agency's own site says "growth marketing for healthcare brands." That is too broad for the buyer question.
Brand-specific prompts reveal another problem. The AI answer can identify the agency, but it cannot explain who the agency is best for. It misses the dental niche and describes the company as a general digital agency. That tells the owner the issue is not only AI. The public positioning is unclear.
The agency chooses three fixes. First, it rewrites the main service page around dental practices, not generic healthcare. Second, it publishes one practical guide on implant lead quality with real workflow examples. Third, it adds a direct audit CTA so a practice owner can move from reading to a conversation.
The next month, the prompt test improves. The agency is still not mentioned everywhere. That is normal. But the brand-specific answers are more accurate, one guide is cited, and AI referral traffic lands on the service page instead of only the homepage. That is useful progress.
Mistakes to avoid
Do not let the test become theater
- Do not run one prompt, like the answer, and call it a strategy.
- Do not change all 20 prompts every month. You need consistency to see movement.
- Do not treat a citation as a guaranteed lead source. It is one visibility signal.
- Do not chase fake mentions, low-quality directories, or thin pages just to influence AI answers.
- Do not ignore traditional SEO data. AI search still depends heavily on crawlable, useful, well-linked web content.
- Do not publish prompt-stuffed content that no real buyer would want to read.
The highest-value fix is usually not a trick. It is making your real expertise easier to understand. Clear service pages, useful examples, buyer FAQs, credible proof, structured internal links, and current information still do most of the work.
Want a practical readout instead of guessing?
If you want to know whether AI search systems understand your business, start with a focused audit. We check crawl access, page clarity, citeability, prompt-test gaps, internal links, and the commercial pages that should carry the visibility work.
Sources
- OpenAI: Publishers and Developers FAQUsed for ChatGPT search discoverability, OAI-SearchBot access, and referral tracking guidance.
- Google Search Central: AI features and your websiteUsed for Google AI Overviews and AI Mode website guidance.
- Google Search Central: Optimizing for generative AI features on Google SearchUsed for Google's current guidance on useful content, RAG, and avoiding AI-only optimization myths.
- Bing Webmaster Tools: AI PerformanceUsed for AI citation, cited page, and grounding-query measurement concepts.
- Google Search Console Help: Impressions, position, and clicksUsed for interpreting traditional search visibility metrics correctly.
- Google Analytics Help: Traffic-source dimensionsUsed for source, medium, referral, and campaign attribution context.
FAQ
How often should an SMB run an AI visibility test?
Monthly is usually enough. Weekly testing can create noise because AI answers vary, and most small businesses do not update their site meaningfully every week. Run the same prompts monthly, record the same fields, and look for repeated patterns.
Which AI tools should I test?
Start with ChatGPT, Perplexity, Gemini, Copilot, and Google AI experiences where available. The exact mix can change, but the principle stays the same: test the surfaces your buyers are likely to use and keep your prompt set stable.
Does appearing in ChatGPT or Perplexity mean my GEO work is successful?
Not by itself. A mention is useful, but you also need accuracy, citations, referral traffic, and commercial outcomes. The better question is whether AI tools understand your business correctly and send the right people to useful pages.
What should I do if competitors appear and my business does not?
Use it as a content and clarity audit. Review the cited competitor pages, then improve your own service pages, examples, FAQs, proof, internal links, and source-ready explanations. Do not copy competitors or chase artificial mentions.
Can I run the 20-prompt test without buying a GEO platform?
Yes. A spreadsheet is enough for the first version. Record prompt, tool, mention, citation, competitor names, answer accuracy, landing page, and next action. Add tools later only if the manual process becomes too slow or too important to manage casually.
