How to Check Your Google AI Visibility in Search Console

What changed

Google introduced the generative AI performance report in Search Console in June 2026 and said on 31 August 2026 that it had rolled out to sites worldwide. The same release added a control that lets a site opt out of generative AI features (sites that opt out receive no traffic or impressions from them). If your property is verified, you can look at your AI impressions today.

What the report shows

• Impressions of your URLs inside generative AI features
• The pages that were shown
• Countries and user devices
• Change by day, week and month
• Generative AI appearances in Google Search and in Discover

An industrial valve manufacturer, for example, can see whether its English product pages appeared in answers in the United States or Vietnam — or whether only the Korean company profile showed up while the product pages were missing.

Three things to read before the total

1. Pages

If only the About page and the blog appear and no product detail pages do, check whether application, specification, certification and delivery details are actually explained in the body text.

2. Countries

If your export markets are the US, Japan and Vietnam but impressions are domestic only, add the product names, standards, industry terms and certifications those buyers use.

3. Trend

Do not draw conclusions from one or two movements. Compare at least four-week periods before and after a change.

What the report does not tell you

• The exact question the user typed
• Actual clicks on your link
• Whether the AI described you positively
• Your recommendation order versus competitors
• Whether any of it turned into an enquiry or a sale

A rise in AI impressions shows improved discoverability. It does not by itself prove recommendation quality or revenue.

Reading four data sources together

A user who discovers you inside an AI answer often searches your company name on Google afterwards, which is recorded as ordinary Google traffic. Adding a "How did you first hear about us? (Google / ChatGPT / Gemini / Perplexity / trade show / referral / other)" field to your enquiry form recovers attribution that analytics tools miss.

Product page improvements for manufacturers and exporters

• Exact product and model names, use cases, target industries
• Materials, specifications, operating conditions, limitations
• Certification and test information, with the supporting documents
• Countries served, minimum order quantity, how to enquire
• Company name, address, contact, last updated date
• Internal links to related products and technical documents

Do not leave the facts inside product photos or a PDF catalogue. Key specifications belong in HTML body text and a clearly headed table, because machines read text far more reliably than images.

Market-by-market use for exporters

If English product pages gain AI impressions in the US but almost none in Indonesia, check local product category names, international and local standards, industry-specific application examples, shipping and lead time, distributor and enquiry routes, and units and currency. One English page for every market is less effective than adding market-specific questions, FAQs and application cases.

A 30-day measurement experiment

1. On day one, record total generative AI impressions, the top ten pages, countries and devices, and whether key product pages appear at all.
2. Choose one important product page with low AI visibility as the test subject.
3. Strengthen the product definition, applications, specification table, certification evidence, FAQ, company information and enquiry route.
4. Log the change date and content, request indexing, then watch the trend for four weeks.
5. Repeat the same questions weekly in ChatGPT, Gemini and Perplexity and record mention and citation rates.

Testing three AI services directly

Comparison only works if the conditions are fixed. Record the date and time, the service and model, whether you were logged in, whether web search was on, the country you connected from, and the language. Repeat each question at least three times, always in a new conversation, to limit personalisation effects.

Common manufacturing prompts:

• Recommend reliable Korean manufacturers of [product].
• Which Korean suppliers export [product] to [target country]?
• Compare Korean [product] manufacturers based on certifications, specifications, and export capabilities.
• What should I check when sourcing [product] for [application]?

Record whether and in what order the company is named; the accuracy of product names, manufacturer-versus-distributor status, export markets and certifications; which page was cited; competitors mentioned alongside; factual errors; and whether the result holds on repetition. Twenty questions × three services × three repetitions is 180 answers a week; comparing four-week moving averages removes most of the noise.

When impressions rise but enquiries do not

• Distinguish being cited as a source from being recommended as a supplier.
• Check whether the AI described your products accurately.
• Check that the link points to the right product page.
• Test that the mobile and English enquiry forms actually work.
• Make sure certificates, test reports and company credentials are on the page.

Checklist

Search Console

• Ownership and the HTTPS / www / subdomain variants are verified
• The generative AI performance report is accessible
• Search and Discover are read separately
• Appearing pages and key export markets are recorded
• Mobile and desktop are compared
• A single-day drop is not treated as a real change

Product pages

• Product and model names appear as text
• Target industries and use cases are explained
• Key specifications are in an HTML table
• Certification and test claims have evidence
• Manufacturer, brand and legal entity names are consistent
• The English page is not raw machine translation
• Export regions and enquiry routes are explicit
• The update date is shown and the mobile enquiry button works
• Anything critical that lived only in a PDF is also in HTML

Direct verification

• The same questions were used across all three services
• Korean and English were tested separately
• Each question was repeated at least three times
• Brand mentions and URL citations were counted separately
• Competitor co-mentions and factual errors were logged
• Conditions were identical week to week

Search Console shows Google’s generative AI impressions, but it does not audit brand mentions in ChatGPT, Gemini or Perplexity, nor the structure of your site. To check the GEO and AEO fundamentals of your own site, run the free AI search optimization audit at geo.ebiz365.com.

Conclusion

The realistic measurement stack is Search Console impressions + web analytics + direct tests in ChatGPT, Gemini and Perplexity + an enquiry-source survey. AI search optimization is not a one-off ranking push; it is the continuous work of measuring where your product pages are found, how AI describes your company, and whether that turns into a real enquiry.

This audit reviews the technical and content structure of a website. It does not guarantee visibility, citation or recommendation in any specific AI service, nor any search ranking.

Sources

• Google Search Central — Introducing Search generative AI performance reports in Search Console (2026-06-03)
• Search Engine Land — AI performance reports rolling out globally (2026-08-31)
• Search Console Help — Generative AI performance report (Search)