The old rule – for the 20 years I’ve been doing SEO – has always been “don’t put text in images”. And now, whilst AI has changed what’s possible, it hasn’t changed what’s best practice.

For years, we’ve told clients:

  • Don’t put headlines in banners
  • Don’t put paragraphs into graphics
  • Don’t turn important content into pretty JPEGs

Because Google’s crawler just sees “image.jpg”.

It couldn’t see the words inside it, so from an SEO point of view, those words basically didn’t exist.

So if your main keyword, message, or call-to-action lived inside an image, you were accidentally hiding it from search engines.

But now AI can read images…

Now, in 2026, things are different.

Google absolutely has the technology now to:

  • Recognise objects in images
  • Read text in images (OCR)
  • Understand context inside visuals
  • Do all the clever stuff you see in Google Lens

So yes — Google can read words inside images now.

Which leads to the obvious question:

“Cool. So can we ignore that old rule now?”

Not quite.

Just because Google can read it, doesn’t mean it treats it like real text

Even though some Google services can read images, Google’s primary way of understanding your page is still:

  • Your HTML text
  • Your headings
  • Your paragraphs
  • Your structure
  • Your metadata

That’s what it trusts.
That’s what it uses to rank pages.

Text pulled out of an image via OCR is:

  • Inconsistent
  • Not guaranteed
  • Harder to interpret
  • Not structured like real page content
  • Not treated with the same weight

So while Google might read the words inside your banner graphic, it does not treat them like a proper H1 or paragraph of content. Which means for SEO purposes, it’s still second-class information.

It’s also an accessibility issue (which Google cares about)

This isn’t just an SEO thing – it’s a usability thing too.Text inside images is invisible to:

  • Screen readers
  • Accessibility tools
  • People using assistive tech

Unless you write perfect alt text (which most people don’t), text in images is basically lost to part of your audience. And accessibility is something search engines increasingly factor into quality signals (for LLMs like ChatGPT too).

Where text in images does get picked up

There are situations where Google absolutely extracts text from images:

  • Photos of signs, menus, posters
  • Infographics
  • Screenshots
  • Social graphics

And you might even see those show up in Google Image Search because Google has read the words.

But that’s very different from: “This page ranks because the keyword was in the banner graphic.”

That’s still not how it works.

So what’s the current best practice?

The best practice is actually the same as before – the same as it always has been.

Do this:

  • Keep important copy as real HTML text
  • Use images for visuals and branding
  • Add proper alt text
  • If you use infographics, include a text version on the page

Avoid this:

  • Headlines as part of images
  • Important messaging that only exists in graphics
  • Replacing text with design for the sake of aesthetics

You still can – and should – have beautiful visuals, just don’t hide your content inside them.

The updated truth about images

The old statement “Google can’t read text in images” is no longer technically true.

The updated version is “Google can read text in images, but it still doesn’t treat it like real page content for SEO.”

And that’s the bit that matters.

The bottom line for images and SEO

AI has changed what’s possible, but it hasn’t changed what’s best practice.

So now we’re telling clients “Google can read it… but don’t rely on it.”

Keep your key content as proper text, using images to support it rather than replace it, and you’ll be absolutely fine.