Google Lens SEO – Image Search SEO: How to Get Your Products Found by Camera

Google Lens SEO – Image Search SEO: How to Get Your Products Found by Camera?
– 1. What Happened?

A shopper sees a product in a café, on a colleague’s desk, in someone else’s shopping bag. They point a phone at it. Within two seconds they have a name, a price range, and a list of places to buy it.

They never typed a word. And whether your brand appeared in that list had almost nothing to do with the keyword strategy your team has been running for the last five years.

Google Lens now handles close to 20 billion visual searches every month… up from roughly 12 billion in mid-2023 and around 3 billion in 2021. Google has said about one in five of those searches is shopping-related. That is roughly four billion commercial visual queries a month, running on a ranking pathway most enterprise SEO services have never audited.

Here is the part that surprises most teams. Google’s own ecommerce documentation is very clear about how products get into Lens results, and it is not primarily an on-site SEO exercise. To be found in Google Lens search results, Google says you need to upload your product details to Google Merchant Center, opt in to product listings, and follow Google’s image best practices.

The opt-in is a setting. Not a ranking factor, not a content strategy, not something you earn… a checkbox in a platform your paid media team probably owns and your SEO team probably cannot see. Every “Google Lens SEO” article circulating at the moment tells you to write better alt text. Very few of them mention that a retailer can do everything else perfectly and still be structurally absent from Lens shopping results because nobody ever clicked it.

That is the gap this article is about.

Google Lens SEO – Image Search SEO: How to Get Your Products Found by Camera?
2. Why Does It Matter?

Visual search collapses the funnel

A typed query is an act of translation. The shopper sees an object, converts it into words, guesses at a brand, types an approximation, and scans results. Every step in that chain is a step your content can intercept.

A camera query removes the chain. The shopper points, and the answer arrives. There is no query to rank for, no long-tail phrasing to capture, no informational stage to own. There is a product, and there is a shortlist.

Your presence on that shortlist is decided before the shopper has articulated a single word.

Lens is not pure computer vision – your page context is a ranking input

This is the most commercially useful thing on Google’s Lens documentation and the most consistently overlooked.

Google describes Lens as comparing objects in the picture to other images and ranking those images by similarity and relevance to the object. That is the visual layer, and it is the part everyone assumes is the whole system.

But Google goes on to say that Lens may also use other helpful signals… words, language, and other metadata on the image’s host site… to determine ranking and relevance.

Read that as a working model:

• Layer one: does this image visually resemble the object in the photo?
• Layer two: does the page hosting this image credibly concern the thing in the photo?

Layer two is ordinary SEO, applied to a surface most teams have never optimised for. The words around your image, the language of the page, the structured data on it and the metadata attached to it all feed a ranking decision that happens without a query.

Your image can be a perfect visual match and still lose to a competitor whose page provides better context about what the object actually is.

Confident matches are winner-takes-all

Google gives an instructive example. When Lens analyses an image, it often generates several possible results and ranks their probable relevance – but it may narrow those possibilities to a single result. Its illustration is a dog judged as probably 95% German shepherd and 5% corgi, where Lens may show only the German shepherd.

Now apply that to your catalogue……

In a text SERP, position four still earns clicks. In a confident visual match, there may be no position four. The distribution is far more brutal at the top than anything your rank-tracking dashboard has prepared you for, and “nearly matched” converts at zero.

Your visual identity is now a retrievable asset

Google also notes that Lens may recognise barcodes or text in an image – a product name, a book title – and return search results for that object.

For brands, this reframes packaging design, label typography, unboxing photography and in-store merchandising as search assets. If a distinctive package is legible to a camera and consistently associated with your domain across the web, it becomes a query. If your product photography is interchangeable with six competitors’, it becomes somebody else’s query.

The demographic skew is a forecast

Google has said the 18-to-24 group engages with Lens the most. That is not a niche audience note – it is the buying population of the next decade telling you which interface they consider default.

Google Lens SEO – Image Search SEO: How to Get Your Products Found by Camera?
3. Who Is Affected?

• Retailers and DTC brands in visually-led categories.

Fashion, footwear, furniture, homeware, beauty, eyewear, plants, food and drink. Anywhere a shopper can see the product in the wild, the camera query already exists. Whether it finds you is the open question.

• Enterprise retailers with a split between SEO and paid media.

This is the structural problem. Lens shopping visibility runs partly through Merchant Center, which usually reports into paid media, while image optimisation sits with SEO and content. Neither team owns the outcome, so neither team measures it. The single most valuable thing many organisations can do this quarter is put both teams in one room for an hour.

• Manufacturers and brands who sell through third parties.

If a shopper photographs your product and Lens surfaces four resellers and not your own site, you have lost a direct-to-consumer sale on your own product. Brands with strong retail distribution and weak visual search presence are effectively funding their resellers’ acquisition.

• Anyone running heavy image CDN transformation.

Responsive delivery that generates a fresh URL for every crop, format and cache-bust is actively working against you. Google asks that an image referenced across a site be consistently referenced with the same URL so it can cache and reuse it rather than re-requesting it. Catalogues serving twelve URLs for one product photo are fragmenting the signal for all twelve.

• Sites rendering imagery via CSS.

Google can find images in the src attribute of an img element, including inside a picture element, but Google does not index CSS images. A gallery built with background-image is not a slow-performing image asset. It is not an asset at all.

• International retailers.

Shopping results in Lens are available in a defined list of markets – the UK, Ireland, most of Western Europe, the US, Canada, Australia, India, Japan, Brazil and others. If you trade outside that list, the shopping pathway is not available to you regardless of how good your markup is.

• B2B and considered-purchase businesses.

Machinery, components, lab equipment, medical devices. A technician photographing a part on a shop floor is a genuine and underserved query type, and almost nobody in these sectors is optimising for it.

Google Lens SEO – Image Search SEO: How to Get Your Products Found by Camera?
4. What Should Businesses Do?

Three workstreams that need to run simultanitolsy. They belong to different teams, and the common failure is running one and assuming it covers the others.

4A. For the marketing team: fix the plumbing first

Nothing else in this article matters if this is not in place. Do this before commissioning a single new photograph.

• Confirm you have a Merchant Centre account with a live, healthy product feed.

Lens shopping results are drawn from product data, and product data reaches Google through Merchant Centre.

• Confirm you have opted in to free product listings.

Google’s ecommerce guidance names this as a distinct requirement alongside uploading the feed. It is a setting in Merchant Centre. Someone needs to physically look at it and confirm its state today — not assume it was ticked at onboarding three years ago, and not take a colleague’s word for it.

• Confirm your feed image quality.

The feed image is frequently the image Lens matches against. Low-resolution, watermarked, heavily-cropped or lifestyle-only feed images will underperform a clean product shot on a plain background. This is a feed asset, not a website asset, and it is often nobody’s job.

• Reconcile the feed against the site.

Prices, availability, titles and identifiers must agree between feed, page and structured data. Where they disagree, Merchant Center can disapprove products, and a disapproved product is not competing anywhere — including Lens. If you have read our pieces on validFrom in offers or invalid SKU values, this is the same discipline applied to a different surface.

• Check your market coverage.

Confirm Lens shopping results are available in the markets you trade in before you build a business case around them.

• Assign an owner.

Not “SEO” and not “paid”. One named person accountable for visual search visibility, with visibility into both Merchant Center and Search Console. This is the recommendation most often ignored and most often the reason nothing changes.

4B. For the content team: give Lens something to read

Layer two of the ranking model is yours. Google uses alt text alongside computer vision algorithms and the content of the page to understand an image’s subject matter – so the surrounding text is not decoration, it is input.

• Write alt text that describes the object, not the keyword.

Google’s own progression is instructive: missing alt text is bad, keyword-stuffed alt text is bad, alt="puppy" is better, and alt="Dalmatian puppy playing fetch" is best. Specific and descriptive beats both empty and stuffed.

Applied to a catalogue: alt="product" is useless, alt="riboflavin vitamin b2 supplement capsules buy online cheap uk" is spam, and alt="Seeking Health Riboflavin 400 mg, 60 vegetarian capsules, white bottle with blue label" is what you want. Describe what a person would see. Colour, material, form, distinguishing features. That is what the camera is looking at.

• Use descriptive filenames.

Google says filenames give light clues about subject matter, and recommends short but descriptive names over generic ones — my-new-black-kitten.jpg beats IMG00023.JPG. For large catalogues, automate the naming from product attributes rather than exporting whatever the photographer’s camera produced. And if you localise images, translate the filenames too.

• Place images next to relevant text.

Google extracts information about an image’s subject matter from the page content, including captions and image titles, and recommends images sit near relevant text on pages relevant to the subject. A product image floating above a wall of generic template copy has weak context. An image sitting beside a specific description of that exact product has strong context.

• Write descriptions that name what things look like.

Most ecommerce copy describes benefits. Visual search rewards copy that also describes appearance — colour, finish, shape, material, size relative to a hand, what it is packaged in. This is the single cheapest content change available and almost nobody does it.

• Check titles and meta descriptions.

Google draws on the title and meta tags when generating the title link and snippet shown alongside image results. The text that persuades someone to tap through from a visual result is the same text you have always been writing – it just now appears in a context you were not planning for.

• Do not neglect category and editorial content.

Google’s ecommerce guidance recommends providing informative product descriptions matching the terms shoppers use, plus category descriptions to cover less specific queries. Lens narrows from many possible matches to few. Contextual depth across a category helps you win the clear ones.

4C. For the development team: make the images indexable

• Use real img elements.

Google finds images in the src attribute of an img element, including when nested inside a picture element, and does not index CSS images. Audit every template for background-image used to display product or editorial photography and convert it.

• Always provide an src fallback.

Some browsers and crawlers do not understand srcset and picture, so Google recommends always specifying a fallback URL in src. Per the HTML standard, a picture element must contain an img fallback.

html

<picture>
  <source type="image/avif" srcset="riboflavin-400mg-front.avif">
  <source type="image/webp" srcset="riboflavin-400mg-front.webp">
  <img src="riboflavin-400mg-front.jpg"
       alt="Seeking Health Riboflavin 400 mg, 60 vegetarian capsules, white bottle with blue label"
       width="1200" height="1200" loading="lazy" decoding="async">
</picture>

• Stabilise your image URLs.

Google asks that an image referenced across multiple pages be consistently referenced with the same URL so it can cache and reuse the image rather than requesting it repeatedly – and flags crawl budget as the reason. In practice:

• Serve one canonical URL per image asset.
• Use srcset for responsive sizes rather than a different base URL per breakpoint.
• Do not append cache-busting version parameters on every deploy.
• Point structured data image properties at the master asset, not a 400px grid thumbnail with query parameters attached.

• Submit an image sitemap.

Google supports image sitemaps for surfacing images it might not otherwise discover, and – unlike regular sitemaps – image sitemaps may include URLs from other domains in the image:loc element, which is what makes CDN-hosted imagery workable. If you use a CDN, Google encourages verifying ownership of the CDN domain in Search Console so crawl errors reach you.

• Use supported formats.

Google Search supports BMP, GIF, JPEG, PNG, WebP, SVG and AVIF referenced in an img src, and recommends the filename extension match the file type.

• Balance quality against weight.

Google is explicit that sharp images appeal to users more than blurry ones and improve the likelihood of traffic from result thumbnails – while also noting images are often the largest contributor to page size. Both halves matter. Aggressive compression that softens product detail is a visual-match liability; unoptimised 4MB hero images are a performance one.

• Nominate a preferred image explicitly.

Google’s image preview selection is automated but can be influenced by supplying primaryImageOfPage, an image property on the main entity, or an og:image meta tag — and advises choosing something relevant and representative, avoiding generic images like your logo or images with text, avoiding extreme aspect ratios, and using high resolution where possible.

• Keep product structured data accurate.

Google notes that structured data makes images eligible for rich results and badges in Google Images, and that the image attribute is required for that eligibility. The same product markup that drives merchant listings supports your image visibility.

• Do not silently opt out.

Google supports opting out of inline linking in Google Images by responding to Google-domain referrers with a 200 or empty 204. Some CDN hotlink-protection configurations do this by default. It is worth confirming yours is not.

4D. Rolling it out across the site

1. Baseline. Pull Google Images impressions and clicks from Search Console (filter Search Appearance to image results). This is the closest available proxy for visual discovery health. Record it before you change anything.
2. Audit the plumbing. Merchant Centre account, feed health, free listings opt-in, feed image quality, market coverage. Half a day. Do it first.
3. Crawl for structural blockers. CSS-rendered product imagery, missing src fallbacks, unsupported formats, missing alt attributes, unstable image URLs. Output a defect count per template.
4. Fix templates before content. One template change corrects thousands of URLs. One alt-text rewrite corrects one. Sequence accordingly.
5. Automate alt text and filenames from product attributes. Brand, product name, variant, format, colour, primary visual feature. Composed from structured fields, not free-typed per SKU. Reserve manual writing for your top revenue lines.
6. Ship the image sitemap and verify CDN domain ownership in Search Console.
7. Re-shoot the top 100 products where photography is the actual constraint. Clean lighting, plain background for at least one asset per product, sharp detail, consistent framing. Feed image and primary PDP image should be the same clean asset.
8. Test manually. Take real photographs of twenty of your own products in realistic conditions — in hand, on a shelf, half-lit — and run them through Lens. Record whether you appear, where, and who beats you. This test takes an afternoon and tells you more than any tool.
9. Monitor quarterly. Image impressions, Merchant Centre product status, and a repeat of the manual test with the same twenty products.

Google Lens SEO – Image Search SEO: How to Get Your Products Found by Camera?
5. What We’re Watching Next

• Multimodal search becoming the default, not the alternative.

Lens is embedded in the Google app, Chrome and Android camera surfaces. Combined with AI Mode’s multimodal input, the camera is becoming a primary query interface rather than a feature you open deliberately. Optimising for it as a side channel will look increasingly odd.

• The Merchant Centre dependency.

More shopping surfaces are being fed from structured product data rather than crawled pages. Organisations where feed quality is a paid-media concern and page quality is an SEO concern will keep finding gaps between the two. We expect feed governance to move formally into technical SEO scope.

• Visual identity becoming a defensible driver.

If Lens narrows confident matches to a single result, product photography and packaging that are genuinely distinctive stop being a branding preference and start being a retrievability advantage. Categories full of white-label lookalikes will find visual search unforgiving.

• Measurement remains the weak point.

There is no Lens-specific report in Search Console. Image-result impressions are a proxy and an imperfect one. Until that changes, manual testing and feed health are the honest measures, and we would rather clients ran a crude test that reflects reality than a precise dashboard that does not.

• Camera-first queries reaching B2B.

Parts identification, equipment lookup and spec retrieval from a photograph are obvious use cases that industrial and technical sellers have almost entirely ignored. The first movers in those categories will have very little competition.

About Szymaniak Digital

Szymaniak Digital is an enterprise AI SEO consultancy working with senior enterprise marketing teams and the engineering teams who ship for them. We audit product data, image infrastructure and structured data at catalogue scale, and design visibility strategies that hold across Google Search, Google Lens, AI Overviews and AI Mode.

If your products are visible when someone types their brand name but invisible when someone points a camera at them, that gap is measurable, and it is usually fixable in a quarter. Book a visual search audit.

Related reading:
Product schema on collection pages; Invalid value in field “sku”; Missing field “validFrom” in offers