For years, SEO conversations have basically started with words.
What are people searching?
What keywords should be on the page?
What questions should we answer?
What should the H1 say?
All sensible questions.
But there is another type of search becoming much harder to ignore.
Someone sees something.
They take a photo.
Circle part of their screen.
Upload an image.
Or right-click something they want to understand.
And search starts from what they can see, not what they know how to describe.
Google has just made that behaviour significantly easier for website owners to measure.
On 24 September 2026, Google introduced multimodal Search reporting inside Search Console. Businesses can now filter performance data specifically for searches involving tools such as Google Lens, Circle to Search, image uploads and Chrome’s “Search this image” feature. (Google for Developers)
That might sound like an SEO nerd update.
I actually think it has much bigger implications for websites.
Because we’ve spent years optimising what our websites say.
We now need to think more carefully about what they show.
People don’t always know what to type
Imagine somebody walks past a restaurant and photographs a dish because they want to know what it is.
Someone sees a piece of machinery and wants to find the manufacturer.
Someone screenshots an interior they love and searches for similar designers.
Someone spots a heritage property and wants to understand its architectural style.
Someone sees a pair of trainers and wants the exact model.
Those are obvious visual-search examples.
But it goes further.
A marketing manager can screenshot a website interaction and search for similar examples.
A founder can photograph packaging and research suppliers.
A facilities team can photograph equipment and look for servicing information.
Someone can circle a product, logo, chart, interface, building, plant, component or piece of furniture without ever knowing its proper name.
Text search requires the person to translate what they see into words.
Visual search removes part of that translation.
That’s incredibly powerful.
Your website images are not just decoration anymore
This is where I think businesses need to rethink how they use imagery.
There are an unbelievable number of websites filled with images that communicate absolutely nothing.
Laptop on desk.
People pointing at Post-it notes.
Random handshake.
Woman laughing at salad.
Man looking very serious while holding an iPad.
You could swap the images between 400 different companies and nobody would notice.
Those images might fill space nicely.
But they don’t explain anything useful about the business.
The more visual search develops, the stronger the case becomes for imagery that actually contains information.
Your product.
Your work.
Your environment.
Your process.
Your team doing the thing you claim they do.
Screenshots of the software.
Before-and-after examples.
Real project photography.
Diagrams.
Plans.
Interfaces.
Results.
Details somebody might genuinely want to investigate.
The question becomes less:
“Does this image make the page look nice?”
and more:
“Does this image help somebody understand what we do?”
That is a much better design question anyway.
This matters for B2B too
Visual search is sometimes treated like an ecommerce feature.
Photograph a shoe.
Find the shoe.
Buy the shoe.
Obviously that use case makes sense.
But B2B businesses shouldn’t assume this has nothing to do with them.
Imagine an architecture practice showing real projects.
An engineering organisation showing equipment.
A recruitment company specialising in surveying showing relevant property and working environments.
A restaurant showing dishes.
A manufacturer showing components.
A software business showing its interface.
A construction business showing materials and finishes.
A web agency showing a distinctive website it built.
Those images contain information.
And when search becomes multimodal, visual information becomes part of discoverability.
This is also why I’d be very cautious about replacing every useful piece of genuine photography with generic AI-generated imagery just because it looks more polished.
If your real work is visually distinctive, show the real work.
Google’s update finally gives us something to measure
This is probably the bit I find most interesting.
Visual search has existed for years.
The problem for marketing teams has been visibility.
If you couldn’t separate that behaviour inside your normal reporting, it was difficult to know whether it mattered to your website.
Google’s new Search Console filter changes that.
If your site receives traffic or visibility through multimodal searches, you can now isolate that activity rather than having it disappear inside wider Search reporting. Google says the data includes Lens, Circle to Search, uploaded-image searches and Chrome’s image-search functionality. (Google for Developers)
That means the first thing I would do is very boring:
Look at the data before reinventing your entire website.
Please do not read this article and immediately commission a “Visual Search Optimisation Strategy™” for £8,000.
Open Search Console.
See whether multimodal activity exists.
Then investigate.
The questions I’d actually ask
If the data starts appearing, I wouldn’t obsess over the total number first.
I’d look for patterns.
Which pages appear?
What types of content are on them?
Are product pages showing up?
Projects?
Articles?
Locations?
Case studies?
Does the visual search activity lead to pages that matter commercially?
Are people clicking?
And most importantly:
What can we learn about what people were trying to identify?
That is useful customer research.
Imagine discovering that a large proportion of visual-search visibility is reaching one particular project page.
I’d want to understand why.
Maybe the architecture is distinctive.
Maybe a diagram is being referenced.
Maybe the product is recognisable.
Maybe Google is associating that image and surrounding page with a topic you hadn’t consciously targeted.
That’s the interesting bit.
Alt text still matters. But visual SEO is not “put keywords in alt tags”
Whenever SEO finds a new thing, somebody inevitably tries to turn it into a checklist hack.
So I can already imagine:
TOP 100 VISUAL SEO ALT TEXT PROMPTS FOR 2027
Please no.
Alt text absolutely matters.
Google says it uses alt text alongside computer vision and page content to understand an image, and useful alt text is also fundamental for accessibility. (Google for Developers)
But Google’s image guidance is much wider than that.
It recommends placing images near relevant text, using meaningful context on the page, using descriptive rather than generic filenames where practical, serving quality images efficiently, using standard HTML image elements so they can be discovered and providing useful metadata. (Google for Developers)
In other words:
The image and the page need to make sense together.
That is not something you solve by stuffing twelve keywords into an alt attribute.
Context is incredibly important
Imagine I upload this image:
IMG_8347.jpg
It shows one of our finished website projects on a phone.
There is no caption.
The surrounding section says:
“Digital excellence without limits.”
Alt text says:
website design london seo webflow wordpress agency web development best web design
Wonderful.
We’ve somehow managed to be vague and spammy at the same time.
Now imagine the same image sits inside a Society of Operations Engineers case study.
The page explains the project.
The screenshot has useful alt text describing the member portal interface.
The surrounding content explains what changed and why.
The project technology is identified.
The page has a useful title.
Internal links connect it to relevant services.
Suddenly both the human and the search engine have context.
That is what optimisation should look like.
Google literally tells us the page around the image matters
This bit is easy to overlook.
Google’s current image-search documentation says that the content and metadata of the page containing an image can significantly influence how and where that image appears in search. It recommends keeping images close to relevant text and placing them on pages relevant to the image’s subject. (Google for Developers)
That means image strategy cannot sit separately from content strategy.
If you’re showing an incredible project but the page barely explains it, you’re leaving useful context on the table.
If your product image sits next to three paragraphs of generic brand copy, same problem.
If a diagram explains something complicated, describe what the diagram is showing.
Help people understand it.
Search gets the benefit too.
Your case studies suddenly become even more interesting
I’ve already been banging on about proper case studies for ages.
This gives me another reason.
A case study can contain:
Project photography.
Interface screenshots.
Before-and-after comparisons.
Process diagrams.
Architecture.
Products.
Brand assets.
Charts.
Results.
Those assets are much richer than another service page containing three stock photos and a gradient orb.
For Raccoon, I’d particularly think about visual project assets in Our Work.
Instead of a case study being treated primarily as written proof, it can become a genuinely searchable visual record of what was built.
That means giving significant project images proper context instead of simply dropping them into a gallery.
What is the screenshot showing?
What problem did that interface solve?
What was changed?
Which part of the experience is important?
Why was that design decision made?
Those details are useful to prospective clients anyway.
Do not hide useful imagery in CSS if it carries meaning
This gets slightly more technical, but it’s worth mentioning.
Google’s image guidance specifically says it discovers images through standard HTML image elements and does not index CSS background images in the same way. (Google for Developers)
That doesn’t mean:
NEVER USE BACKGROUND IMAGES.
Decorative visual backgrounds are completely fine.
But if an image contains meaningful content you want users and search systems to understand, it should be implemented appropriately.
This is one of those tiny development choices that becomes important later.
The image might look identical on screen.
Underneath, however, one implementation actually communicates what the asset is.
Good front-end development is full of little things like this.
High quality does not mean upload a 14MB JPEG
There is another trap.
If visual search means imagery becomes more important, someone will hear:
“USE HIGH-QUALITY IMAGES.”
and upload a photograph directly from a professional camera at roughly the size of a small moon.
Please don’t.
Google recommends strong image quality while also warning that images are often one of the largest contributors to page weight. Responsive image techniques and proper optimisation matter because visual quality and website performance need to coexist. (Google for Developers)
Your beautiful photography is not helping conversion if the user leaves before it loads.
This is where the platform implementation matters.
Webflow, for example, automatically creates responsive image variants for normally uploaded supported images so browsers can request a more appropriate size for the device rather than always downloading the full original. (Webflow Help Center)
WordPress has its own responsive-image handling.
Other platforms have their own systems.
Use them properly.
Your Open Graph image should not be an afterthought either
This is one of those launch tasks that somehow always ends up being done at 5:47pm.
“What image should show when somebody shares this page?”
“I don’t know. Use the logo?”
Google recommends using a relevant, representative, high-resolution image for page-level image metadata rather than simply defaulting to generic imagery such as a logo. (Google for Developers)
That makes sense beyond search.
Your social card.
Your shared Slack preview.
Your LinkedIn link.
Potential search appearances.
All of them benefit from an image that actually represents the page.
I’d start treating the primary page image as part of the content brief rather than something somebody remembers during launch QA.
AI imagery needs more thought than “does it look realistic?”
I’m not anti-AI imagery.
I’ve used it.
It can be incredibly useful.
The problem is when businesses start using generated imagery to replace evidence.
If your page explains your actual restaurant, show your actual food.
If you’re a hotel, show the rooms people will stay in.
If you’re an architect, show the work.
If you’re selling a physical product, show the product accurately.
If you’re a software company, show the interface.
If you’re an agency, show what you built.
Generated visuals are great when the purpose is illustrative, conceptual or creative.
But there is commercial value in visual truth.
Your website should not look more impressive than the thing a customer actually receives.
That isn’t an SEO principle.
It’s a trust principle.
And trust converts.
Marketing teams need an image library that contains meaning
This is where operations comes into the conversation.
A lot of businesses have an image library that looks like this:
final.jpg
final2.jpg
new-final.jpg
DSC_8834.jpg
homepage-new-use-this-one.jpg
And 400 mystery images inside WordPress.
If visual assets are becoming more important for discovery, accessibility, content and conversion, asset management deserves slightly more respect.
For important images, the team should know what the asset is, where it is used, whether it has appropriate alt text, whether there is a higher-resolution original, whether it can legally be reused, and whether it still represents the business accurately.
That does not require an enterprise DAM implementation.
It requires not treating your website media library like a kitchen drawer full of batteries and takeaway menus.
Every platform is making this easier
Modern website platforms are increasingly putting this kind of work closer to the content team.
Webflow lets teams define descriptive or decorative images and now uses its AI tooling to help identify missing alt text across a site, although its own documentation notes that some images inside components and CMS-bound content still need manual review. (Webflow Help Center)
Framer’s plugin ecosystem is also filling out around pre-publish QA and image auditing. A Framer plugin released on 19 September, for example, checks things such as missing alt text, heavy images, broken links and incomplete CMS content before publishing. (Framer)
And WordPress gives developers and site owners huge flexibility over image markup, media handling, responsive images and structured data depending on the theme and stack being used.
The platform isn’t the strategy.
But good tooling makes a good process much easier to maintain.
Here is the visual-search audit I’d actually run
I would not redesign anything yet.
First, open Search Console and use the new multimodal search filter if data is available.
Look at the pages receiving impressions or clicks.
Then take your most commercially important pages and look at every meaningful image through four questions.
Does this image show something worth understanding?
If not, maybe it is purely decorative. That’s fine. Not every image needs to become an SEO asset.
Would somebody know what it means without the rest of the design?
If the answer is no, check whether the surrounding copy, caption or alt text supplies the missing context.
Is this genuinely ours?
Real work, real people, real products and real places create far stronger evidence than interchangeable stock imagery.
If someone discovered us through this visual context, would the landing page help them take the next step?
Because discovery without conversion is still just discovery.
That last question matters.
Visual discovery still needs good CRO
Imagine someone uses Lens to identify a product.
Your website appears.
Brilliant.
They click.
The product is out of stock and there are no alternatives.
Or the page gives them no price.
Or the contact button doesn’t work.
Or the page is unusable on mobile.
Visual search did its job.
The website didn’t.
The same rule applies to every acquisition channel.
SEO cannot rescue bad UX.
Paid ads cannot rescue bad UX.
AI recommendations cannot rescue bad UX.
Visual search cannot rescue bad UX.
The page receiving the visitor still needs to answer their question and create an obvious next step.
This gives marketers another source of customer language
There is also a slightly less obvious benefit.
Multimodal search data can tell you about things people struggle to describe.
Traditional keyword research shows you the words someone used.
Visual-search behaviour can reveal the objects, styles, products and visual characteristics behind the question.
That could become brilliant research material.
A property company may discover users are visually searching particular architectural details.
A food business may find one dish driving disproportionate discovery.
A manufacturer could identify components being searched visually.
A designer might see a specific style or project repeatedly surfaced.
The goal isn’t simply:
“How can we rank for this image?”
It’s:
“What does this tell us our audience is interested in?”
That’s where marketing gets more interesting.
Search is becoming less about knowing the correct words
This is the wider shift.
Text search assumed a user could articulate what they wanted.
Multimodal search means the input can be:
Words.
Images.
A camera.
A screenshot.
Something circled on a screen.
Potentially combinations of them.
That doesn’t make written content less important.
Quite the opposite.
Search systems still need context around what they see.
But websites need to communicate through more than one medium.
The copy explains.
The image proves.
The structure connects them.
The metadata adds context.
The page experience converts the interest.
That is a proper website strategy.
Final thoughts
I don’t think businesses need to suddenly rename image files for three days and call it “multimodal SEO”.
That misses the point.
Google giving website owners dedicated multimodal-search reporting this week is interesting because it makes another part of the customer journey visible. (Google for Developers)
And it highlights something websites should probably have been doing properly anyway.
Use meaningful imagery.
Show the actual work.
Give images useful context.
Make them accessible.
Optimise them properly.
Build pages that help people understand what they’re looking at.
Then measure whether people are finding them.
We’ve spent years asking:
“What would somebody type into Google to find us?”
I think there’s now another question worth adding:
“What might somebody see that leads them to us?”
That is a very different search journey.
And now, finally, we can start measuring it.
Discover more from Raccoon Web
Subscribe to get the latest posts sent to your email.
