Google Search now includes AI Overviews, AI Mode, and multimodal search experiences. The measurement job is not to invent a single “GEO score.” It is to separate what Search Console can actually observe from what still needs controlled cross-platform measurement.
Google introduced dedicated Generative AI performance reporting in Search Console in June 2026 and said the insights rolled out globally by August 31. On September 24, Google added reporting for web multimodal search, including searches that use Lens, Circle to Search, image uploads to Google Search, and Chrome’s image-search action. These additions make first-party measurement more useful, but they do not turn Search Console into a universal AI-citation tracker.
Use the dedicated Search Console views to understand when pages appear in Google’s generative AI features and inspect pages, dates, countries, and devices where available.
Use the multimodal search filter when your site receives eligible activity. This is especially relevant for businesses whose images, products, places, diagrams, or visual examples help answer a search.
Keep Page × Query analysis. AI features do not make ordinary query ownership, indexing, snippets, internal links, or conversion measurement obsolete.
| Layer | Question | Best evidence |
|---|---|---|
| Eligibility | Can Google crawl, index, and serve the page? | URL Inspection, robots, canonical, indexability |
| Search visibility | Which pages and queries are surfacing? | Search Console Page, Query, Page × Query |
| Google AI visibility | Is the site appearing in Google generative AI features? | Search Console Generative AI reports |
| Visual discovery | Is content surfacing through image-led search? | Multimodal search reporting |
| Site behavior | What happens after a search visit? | GA4 landing pages and meaningful events |
| Business outcome | Did visibility produce a qualified opportunity? | Lead and sales records with durable attribution |
A sitewide average can hide the exact search ownership problem you need to solve. Export page and query data, then ask which URL owns each important intent. A service page should not be judged the same way as an informational guide. For local service businesses, split commercial queries, problem queries, branded searches, and geographic modifiers before deciding what changed.
StartLab’s Search Feedback Loop is useful here: search language becomes an evidence source for page decisions, not a reason to publish every phrase as a new URL.
Google’s generative AI reports can show visibility inside Google’s own AI search experiences. That is valuable first-party evidence. It still does not prove that a brand was recommended in every answer engine, nor does it tell you that an impression created a qualified lead.
Keep a separate ledger for direct observations from other AI platforms if those platforms matter to the buying journey. Record the prompt set, date, platform, brand mention, linked source, and variability across repeated tests. Do not blend that dataset into Search Console and call the result one universal rank.
The September 2026 multimodal reporting update matters because modern discovery is not always typed text. A customer may photograph an object, upload an image, or use a camera-based search flow. If your business depends on visible products, locations, repairs, before-and-after evidence, diagrams, or field examples, image quality and contextual relevance deserve their own measurement lane.
Use descriptive alt text for accessibility and context, keep important explanatory information in text, and make sure the page around the image clearly explains what the visual represents. Do not create decorative image volume simply to chase a new report.
An AI-feature impression is not a lead. A click is not a customer. Combine Search Console with analytics and a durable lead record. At minimum, preserve landing page, source context, meaningful website action, qualification state, and downstream outcome where your systems can support them.
If impressions rise but clicks do not, use the existing StartLab click-gap diagnostic before rewriting pages. SERP features, intent, ranking distribution, snippets, local results, and AI answers can all change click opportunity.
Check newly published and commercially important URLs for indexability, canonical consistency, and crawl problems.
Compare Page × Query and generative-AI visibility by a consistent period. Flag meaningful changes, not one-day noise.
Review landing-page behavior and qualified actions. Keep “visibility improved” separate from “business outcome improved.”
A single manually tested prompt is not a stable ranking. A Search Console impression is not proof of a recommendation. A third-party visibility score is not revenue. And a page being technically eligible does not guarantee that Google will crawl, index, or surface it. The useful discipline is to label each observation by what it actually proves.
For changing Google behavior, use the source documentation: Google’s Generative AI performance report announcement, Google’s multimodal Search Console announcement, and Google’s AI features guidance.
If your Search Console data is growing but the business meaning is unclear, StartLab can help separate technical eligibility, query ownership, AI-search visibility, click opportunity, and qualified outcomes.
Run the StartLab Growth Audit Explore SEO & AI Search Services