Visual search lets people find products from a photograph, not just from words. If you sell objects with recognisable features, you need your own photographs, clear product pages and accessible pages so Google can interpret your catalogue. Search Console now includes a filter for multimodal searches, but appearing in those results is not guaranteed.
What changes when a customer searches with an image?
Someone sees a lamp in a shop window, photographs a texture or spots a handbag in a magazine. They may not know how to describe it: the exact material, the style name or the brand. Their question starts with what they see. Visual search tries to match that reference with content that search engines can interpret.
This affects decoration, fashion, craft, furniture and products with aesthetic variations in particular. It can also help an industrial supplier when a part is recognised by its shape better than by its trade name. It does not mean every business will win sales this way. It means the photograph is now part of the commercial information.
Google announced on 24 September 2026 a new performance filter for multimodal web searches in Search Console. According to its documentation, it groups visual queries coming from experiences such as Lens, Circle to Search, image uploads and image search from Chrome. The rollout began globally that day and metrics appear once the property receives that kind of traffic. Source: Google Search Central, 24 September 2026.
As a hypothetical example, think of a ceramics shop selling a mug with an uneven blue glaze. A customer might search "handmade blue mug", but could also show a photo of a similar object. If the product page only says "model 47" and shows one small photograph, it gives little context. An honest description of the finish, dimensions and variations helps the buyer, even if it does not ensure a place in Lens.
How do you prepare photographs that help sell?
The main photograph must make clear what is being sold. An artistic image can inspire, but if it hides the shape, size or finish, it leaves questions open. A useful catalogue combines a clean shot of the object with other images showing scale, details and real use.
Do not confuse optimisation with applying an AI filter to every photo. False reflections, exaggerated colours or altered proportions create expectations the product cannot meet. For items with textural variation, a faithful photo can be commercially more valuable than a spectacular scene.
A practical plan starts by choosing one representative product and reviewing its assets: front view, side view if it adds information, a material detail and a use-in-context shot. Composition changes by sector. A shoe needs its profile and sole; a tile, texture and scale; a tool, its connections. There is no universal visual template.
On the website, files should load without blocking the page. Google's image documentation explains that it uses several signals to understand images, including the page's surrounding text and the alt text. Accessibility matters too: useful alt text describes the relevant content, not a pile of keywords. Source: Google Search Central documentation, accessed 14 October 2026.
The link between image and variant deserves attention. If a product page lets people choose red, green or black but always shows the same photo, a buyer may purchase with the wrong expectation. Match each choice to its photograph and check that the variant is still available. Trust begins before payment.
What information does the product page need?
The image identifies; the page explains. A page that wants to attract useful visitors must quickly answer what the product is, who it is for and what its limits are. A recognisable name, verifiable features, availability and terms weigh more in the buying experience than a paragraph full of adjectives.
For a shop with visually similar items, differentiate materials, measurements, compatibility and care. If two chairs look identical but only one is suitable for outdoor use, that difference must be written. Do not expect a system to deduce from a photograph a property that cannot be seen.
Product structured data can help systems understand certain attributes, but it does not replace visible content or guarantee a rich result. Google documents requirements and recommendations for product data; the marked-up information must match what people see. Source: Google Search Central documentation, accessed 14 October 2026.
A well-built product page also has clear navigation. If a buyer finds an interesting photo but lands on a generic category with hundreds of items, the opportunity fades. Review product URLs, stock status and whether people can get from search to a specific decision without unnecessary steps.
Alt text is not an advertising label. "Cylindrical cobalt-blue vase with uneven glaze on a white background" says more than "best cheap vase buy online". Describe what actually appears. If an image is decorative and adds no information, you do not need to force an SEO description.
How do you read the new Search Console report?
Start by checking that your property is verified and that you can open the performance report. Google says the multimodal search-type filter is rolling out and that data will appear only if your site receives traffic attributable to those queries. A missing line does not prove a technical fault. There may not be enough activity yet, or the report may not be available for that property.
If data appears, analyse pages and trends before drawing conclusions. A rise in impressions is not an increase in orders. A fall may reflect changes in demand, seasonality or coverage. To judge the business, cross-check against orders and enquiries that actually reach your systems, within the limits of attribution.
Do not automatically add the new filter's clicks to other reports as if they were unique visitors. Search reports classify interactions in specific ways; your web analytics measures sessions and events under different rules. Define what each counter compares before taking it into a sales presentation.
A hypothetical example: a shop finds that three tableware product pages account for most multimodal impressions. The reasonable decision is not to redesign the whole catalogue. First it would review those three pages: photograph, variants, stock, visible price, description and ease of buying. Then it would compare behaviour with equivalent pages.
It is also worth recording the dates of your changes. If you renew photographs, text and template on the same day, you will not know afterwards which change helped. Work in small batches and note your hypotheses. A measurable improvement needs an earlier reference point, even if that reference is not perfect.
What should you test before investing in more images?
Audit ten product pages, not a thousand. Choose products with known demand and check on a phone how their images look, how long they take to load and whether the important information is available without guessing. A good test starts from a specific problem, not from a catalogue of tools.
You can use this sequence as an editorial check:
- Recognition: does the main photo represent exactly the item and its variant?
- Context: can people understand size, material and use without assumptions?
- Access: are the image and the page reachable on the public website?
- Purchase: do availability and terms match your inventory?
- Measurement: do you know which commercial result you want to observe?
Before commissioning a large photo shoot, estimate the effort of maintaining the catalogue: who updates images when a finish changes, who removes discontinued products and who checks that variants are linked correctly. Quality is not a one-day job but an operational responsibility.
How do we approach this at meiika?
At meiika we approach websites and digital growth through the whole journey: how an offer is presented, what information the buyer needs and what happens when they want to get in touch or buy. A visual improvement only makes sense if it connects to a usable, maintainable website. You can see our work on business websites and in Growth & MarTech.
We do not promise appearances in Google Lens, nor do we attribute sales to a new format without measuring them. We prefer to review the product pages, images and data that already exist, find the blockers and define a test proportionate to the business.