What's the best way to test whether a new comparison page gets my product into AI answers?

The best test is a small, controlled before-and-after experiment—not one-off prompting. AI answers vary, so track the same realistic buyer questions repeatedly and compare the page against a baseline.

1. Choose 10–20 non-branded comparison questions buyers might ask, such as “What are good alternatives to [competitor] for [use case]?” Include several relevant competitors.
2. Record a baseline before publishing: test those exact questions on the AI services that matter to your customers. Save the answers, dates, and any cited URLs.
3. Publish the page, then repeat the same test on a regular schedule—say weekly for 4–6 weeks. Keep the wording, location, account state, and other test conditions as consistent as possible.
4. Track more than whether your brand appears:
   • Was your product named or recommended?
   • Was the new page cited or linked?
   • Was it described accurately?
   • Did competitors appear instead?
5. Use a comparison group if you can. Track similar questions that the page was not designed to answer. If visibility rises only for the questions the page targets, that’s more suggestive than a general increase across all questions.
6. Check whether the page is crawlable and indexed, and watch analytics for AI referrals. OpenAI says ChatGPT referral URLs include utm_source=chatgpt.com; referral clicks are useful evidence, but they won’t capture answers that mention you without sending a visit. (help.openai.com)
7. Treat Google separately. Google’s AI features follow its general Search eligibility and SEO guidance; Search Console can help assess Search performance, but don’t treat impressions or clicks alone as proof that this specific page caused an AI answer to mention you. (developers.google.com)

A simple scorecard is share of tested answers that mention you, share that cite the new page, and accuracy of the product description—reported separately for each AI service. Save the actual answers as evidence; an aggregate “AI visibility” score can hide what changed. The key caveat: this shows a pattern, not definitive causation, because answers can vary with time, query, and retrieval context.

The most reliable approach is a repeated, before-and-after test on a fixed set of prompts, not a single check.

1. Set a baseline before you publish. Pick 15–20 buyer-intent prompts that the new page should answer, such as "best [category] tools" or "[your product] vs [competitor]." One guide suggests testing 15–20 buying-intent prompts across ChatGPT, Perplexity, and Gemini, then recording mentions for your brand and competitors and calculating each brand's mention rate. Run them now, so you have a "before" number.

2. Run each prompt several times. Answers vary between runs, so one screenshot proves little. One 2026 guide warns that single-run measurements can create misleading precision. Repeat each prompt and report a rate, such as the share of runs where you're mentioned.

3. Use clean sessions and turn on web search. Start with a clean session to avoid prior context shaping responses, using incognito mode or logging out. In ChatGPT, one method is to run each prompt with web search enabled, then check the answer body for your name and the source list for your link.

4. Log more than yes/no. Track the query, platform, date, whether you're mentioned, your position, how you're described, which competitors appear, and which sources are cited. "Cited" means the engine links to you, while "mentioned" means your name appears in the answer text without necessarily linking. Citation depends on the engine being able to reach, parse, and lift a clean passage from your page. Those are separate outcomes, so record them separately.

5. Check whether the new page is the source. Perplexity usually shows source links, but ChatGPT does not always display direct citations, and your brand may appear without a clickable source link. So a mention doesn't prove the page drove it. Look for your page's URL or distinctive wording where you can.

6. Use a control. Compare the prompts you target with similar comparison prompts the page doesn't cover, and compare against a competitor's comparison pages. This helps separate the page's effect from general brand awareness. This is my suggestion, not something from the sources.

7. Watch referral traffic too. For the traffic side, the most reliable method is tracking AI referral traffic in Google Analytics 4, filtering for chatgpt.com, perplexity.ai, and gemini.google.com.

8. Allow time, then retest. Rerun the same prompt set on a regular cadence, such as monthly, and compare against the baseline. I don't have a source for a specific waiting period, so start with a few weeks and adjust based on what you see.

Tools: If you need to scale, GEO trackers like Ahrefs Brand Radar or Semrush's AI tools run prompts across platforms and report mention and citation rates. A manual test of 20 prompts across three platforms is enough to start.

What to expect: A 2026 study of 100+ brands found that the ranked "best-of" listicle is the most-cited content format, at about 21% of all citations. A comparison page is a reasonable format to test. That study was run by a GEO vendor, so treat the figures as directional.

To test whether your new comparison page is successfully appearing in AI answers, you should move beyond traditional rank tracking and adopt a multi-layered approach that monitors both visibility (impressions) and direct citation.

1. The Primary Diagnostic: Google Search Console

As of late 2026, Google Search Console is the most reliable "official" way to track your performance in generative AI features.

• Use the Generative AI Report: Navigate to the Performance tab in Search Console and look for the Generative AI report (often broken down into Search/AI Overviews and Discover). This will show you which of your URLs are appearing in AI-generated responses and the volume of impressions they are generating.
• Analyze Trends: Since this report tracks impressions (not necessarily clicks), look for upward trends in impressions for your comparison page following its publication. If the report is missing or shows zero, it may indicate that your page has not yet gained enough visibility to be reported or lacks sufficient authority for those specific queries.

2. Manual Verification (The "Human" Audit)

Automated tools are helpful, but manual testing is essential to see the "why" behind the results.

• Define Your "High-Intent" Queries: Identify 10–15 specific questions your customers actually ask, such as "What is the best [Product Category] for [Specific Use Case]?" or "Compare [Your Product] vs. [Competitor]."
• Search in "Clean" Environments: Perform these searches in incognito/private windows or using specialized SEO tools that simulate user locations.
• Check Different Engines: Don't limit yourself to Google. Test the same prompts in ChatGPT, Perplexity, and Claude. These models often pull from different indexes, and seeing your site cited in one but not others can reveal where your content structure is (or isn't) working.

3. Use Dedicated AI Visibility Tools

Manual checks are inconsistent and time-consuming. Several specialized tools can help you track citations systematically:

• AI Visibility Trackers: Tools like Semrush (AIO Visibility Checker) or AEO Engine can automate the process of querying AI models for your target keywords and reporting whether your page is cited, which competitors are appearing, and if you are being left out.
• What to Look For: Use these tools to capture the actual AI response. A "no citation" result isn't just a failure—it’s a data point. Check if the AI provided a different answer entirely or cited a competitor, which helps you understand what content (e.g., specific features, pricing tables) you might be missing.

4. Optimize for "Extractability"

If you find your page isn't being cited, it is likely because the AI cannot easily extract the answer. To improve your odds:

• Direct Answers: AI-cited pages often answer the user's question in the very first few lines of the page.
• Structure Matters: Use tables, lists, and clearly defined "Pros vs. Cons" sections. AI models are significantly more likely to cite content that is already formatted for quick extraction rather than dense, narrative prose.
• Align with Intent: Ensure your page explicitly addresses the "versus" or "best-for" intent. Data suggests that "X vs Y" comparison pages are currently the strongest predictors of AI search traffic in B2B and commercial sectors.

5. Measure "Halo Effect" Traffic

Since AI answers are often "zero-click" (users get the answer without visiting your site), traditional traffic metrics may be misleading.

• Watch Branded Search: If your page is being cited in AI answers, you may see a "halo effect"—an increase in direct traffic or branded search queries (people searching for your company name by heart) shortly after the AI answer appears.
• Check GA4: While traffic from AI is often lumped into "Organic Search," keep an eye on your conversion rates. AI-referred traffic often has a lower bounce rate and higher engagement because the user has already received a "recommendation" from the AI before clicking through.

The best test is a repeatable before-and-after check, not a one-off prompt. AI answers vary between runs, so look for a pattern across questions, engines, and time.

1. Set a baseline before launch. Write down 10–20 realistic buyer questions that might lead to your comparison page, such as “Product A vs Product B” and “best [category] for [use case].” Run the same prompts in the AI products your customers use.
2. Repeat after publishing. Use the same prompts in fresh chats, across multiple engines, and run each prompt a few times. Check again after a few weeks; a single answer can be noisy. ⁶⁷
3. Record separate outcomes:
   • Is your product mentioned—and where in the answer?
   • Is the new page’s exact URL linked as a source?
   • Which competitors and other sources appear?
   • Is your product described accurately and favorably?
4. Compare rates, not anecdotes. For example, track the share of runs that mention your product and the share that cite the new page. Keep those as distinct measures: a brand mention isn’t the same as a page citation. ⁶¹¹
5. Check traffic and business impact too. In Google Search Console, AI-feature links are included in Search performance reporting; Google’s documentation also describes a generative-AI performance report for impressions and pages, where available. Use analytics to look for visits and conversions from AI referrals as a downstream signal. ¹⁴

A simple spreadsheet with date, engine, prompt, product mentioned, answer position, page cited, competitors cited, and referral/conversion data is enough to start. If the page isn’t cited, that doesn’t necessarily mean it had no effect: AI answers can mention a product without linking to the page, and citation sources can shift from run to run. ⁵¹²