Google Lens and AI Search: Measure Visual Product Discovery
Google's new Search Console filter makes image-led search measurable. Use this Lens audit to separate impressions, clicks, AI exposure, and sales.
A buyer takes a photo of a lamp, draws a circle around a shoe or uploads an image to find a similar product. Your keyword report may miss the starting point entirely. Google's new multimodal Search Console reporting makes this behavior more visible—but it does not turn every image impression into a purchase.
The opportunity is to build a visual discovery audit: which pages receive image-led exposure, what products those pages actually show, and whether the journey reaches a suitable offer.
- Use the new filter as evidence, not as a complete buyer diary. It covers specified Google web multimodal searches and is rolling out globally from September 24, 2026.
- Pair query data with an image-and-page audit. A good photo on the wrong variant page is a poor buying experience.
- Keep visual exposure, AI answer appearance, visit and order separate. The first does not establish the last.
Methodology & sources
Editorial review for factual claims (as of 2026-09-28).
We reviewed Google's September 24 Search Console announcement, Image SEO guidance and Product structured-data documentation. The sample workflow below is ours. Google says the reporting rolls out globally starting on that date and appears when the site receives traffic from those queries; we have not inferred that every property already has the filter.
What the new data does—and does not—cover
Google says the multimodal filter appears in both Search-results performance and its generative-AI performance reporting. It includes Google Lens, Circle to Search on Android, image uploads to Google Search and Chrome's “Search this image” action. Export the data, select your product-page group and compare it with the prior period. Segment by country and device where the interface offers those dimensions; keep the report definition attached to each export.
Do not relabel the filter “all Lens sales.” It is Google's reporting for specified web multimodal searches, not a cross-platform total, a transcript of the image the person submitted, or proof that a displayed page was recommended by an AI answer. If the filter is not visible yet, record that as report availability, not zero visual demand.
Inspect the journey, not just the picture
Pick 20 representative products: best sellers, visually distinctive items and products often confused with another variant. Review each across four steps:
- Image: Clear primary photo, accurate color and proportion, useful alternate views. Do not make a generic lifestyle photo the only evidence of a specific variant.
- Context: Descriptive filename and alt text, with surrounding copy that names the actual product and use case. Google says filenames offer a light clue, while alt text and relevant nearby text matter for understanding and accessibility.
- Landing page: The image should lead to an indexable, usable page for the correct product or category—not a discontinued SKU or a page with contradictory stock data.
- Commercial truth: Variant, price, shipping, returns and compatibility should survive the move from discovery to checkout.
| Observation | Possible interpretation | Next check |
|---|---|---|
| Multimodal impressions rise; visits stay flat | More exposure, unclear buyer intent or presentation | Compare landing-page relevance and click trend |
| Visits rise; orders do not | Product mismatch, stock or price friction | Check variant, checkout and returns |
| AI-feature exposure rises | Google surfaced a URL in an AI feature | Inspect the actual answer before claiming recommendation |
These are hypotheses, not automated diagnoses. Keep an annotated before-and-after record when replacing images or rewriting attributes. A change in Google reporting can reflect rollout, seasonality or demand as well as your work.
A useful experiment for a retailer
For one category, fix 10 pages with ambiguous images and unclear variant copy. Leave a comparable category unchanged during the same period, if practical. Record image changes and availability before comparing multimodal impressions, visits and purchase quality. This is a directional test, not randomized causal proof. Ask a few real buyer questions with a photo in the target market and save the observed result—but do not replace aggregate Search Console evidence with one screenshot.
The best question is not “Did Lens find us once?” It is “When a shopper starts with an image, do they reach the right, buyable item—and can we verify that path?”
Frequently asked questions
The same Q&A pairs ship as FAQPage structured data so AI engines can quote them verbatim.
- Can Search Console now show Google Lens and other image-led searches?
- Google announced a multimodal search filter for Search-results and generative-AI performance reports on September 24, 2026. It covers specified web searches including Lens, Circle to Search, image uploads and a Chrome image-search action. The global rollout is gradual and data appears when a property has qualifying activity.
- Do multimodal impressions mean that AI recommended my product?
- No. An impression means a URL appeared under Google’s reporting rules. It does not establish that an answer recommended the brand, that the shopper clicked, or that an order followed. Inspect the customer-facing result and keep visits and sales as separate verified outcomes.
- What should a retailer fix first for visual search?
- Start with accurate variant photos, descriptive alternative text, useful surrounding copy and a landing page for the right available product. Compare price, stock, shipping and returns with the checkout. This makes the image-led journey more useful to buyers, but it does not guarantee ranking or a sale.
Primary sources and reporting limits
- Google Search Central: web multimodal performance reporting — surfaces, filters and rollout.
- Google Search Central: Image SEO best practices — filenames, alt text and page context.
- Google Search Central: Product structured data — commercial details and product eligibility.
Reviewed September 28, 2026. Filter visibility depends on rollout and qualifying activity; absence of a metric is not proof of absent demand.
Related articles
Your GEO Score
Establish an AI mention baseline you can defend
GEO Tracker AI runs repeatable checks for supported engines so you can see whether your brand is mentioned, what context shows up, and how that changes week over week — complementary to Search Console, not a replacement for it.