Hey – it’s Fiona!
Last week we talked about why your next user might not be human. Today, let’s zoom in on the practical side: how AI agents actually “read” the web.
This is one of those topics that sounds abstract until you see it in action, then suddenly it clicks. Once you understand what agents see (and what they skip), you’ll spot blind spots in your own designs straight away.
The surprising thing agents don’t do
Humans browse visually. We scroll, skim, and take cues from colour, layout, and design. We’ll forgive a vague headline if the imagery feels right.
AI agents? They don’t care about any of that. They:
Strip away your design.
Read your code, not your vibe.
Focus only on information they can extract, structure, and verify.
That means your gorgeous hero video, clever copy, and custom icons? Invisible.
Four steps of an agent’s visit
When an AI agent “lands” on your site, here’s what happens:
1. Crawl
Agents find your pages through links, sitemaps, APIs, or search indexes.
→ If your site isn’t linked or mapped clearly, it might as well not exist.
2. Parse
They scan your HTML for structure:
Headings (H1, H2, H3)
Metadata (titles, descriptions, Open Graph)
Structured data (schema markup)
Alt text for images
Internal linking patterns
Design elements? Ignored.
3. Extract
They pull out the hard facts:
Product or service names
Features and benefits
Prices and availability
Location and contact details
Author and publish date for articles
If a detail isn’t in plain text, they won’t see it.
4. Compare
Finally, they line you up against competitors:
Is the information complete?
Is it trustworthy and consistent?
Does it match the human’s request exactly?
This is the moment you either make the shortlist — or disappear.
What they ignore (and why it matters)
Here are common elements humans love but agents completely disregard:
Background videos or images (without captions/alt text)
Playful headlines that don’t state the obvious
Key facts hidden in PDFs, graphics, or gated content
Inconsistent business details (different phone number on site vs. Google profile)
For us, these might be nice creative choices. For agents, they’re just black holes.
Seeing through two lenses
One of my favourite exercises: take a page you’ve designed and view it through both “human” and “agent” lenses.
Human lens: Does the design feel credible? Is the story clear? Does it guide me to take action?
Agent lens: Could I strip this back to text and still get the facts? Are the headings clear? Is schema markup in place?
When you do this, you’ll often find that what delights humans leaves machines confused, and vice versa.
A real-world example
I ran a quick audit of a service business site recently. Humans would love it: bold imagery, emotional storytelling, clever copy. But through an agent’s eyes, here’s what happened:
Crawl: ✅ All pages linked in the sitemap.
Parse: ⚠️ Headings were inconsistent, with multiple H1s per page.
Extract: ❌ Pricing hidden in a downloadable PDF. No schema markup.
Compare: ❌ Key service descriptions too vague (“solutions for modern teams”).
Result? Invisible in agent-curated recommendations — despite looking great to people.
After a few small fixes (plain-text pricing, clearer headings, consistent schema), the same business started showing up in AI-generated shortlists.
Why this matters now
This isn’t just “SEO 2.0.” It’s the beginning of AI-first discovery. If you’re not machine-readable, you risk being invisible.
And here’s the thing: most of your competitors aren’t optimising for agents yet. Which means even small improvements (like adding descriptive alt text or fixing schema) can give you an outsized advantage.
Where to start
Don’t overcomplicate this. Pick your homepage or one key product/service page and:
Strip it back to raw text. Could an AI agent get the essentials?
Check for the basics: one H1, factual headings, alt text, schema, consistent details.
Rewrite any vague copy into plain text facts.
These are tiny changes with huge impact.
Try this this week
Open your site in “Reader View” (most browsers have this) or run it through a basic text-only tool.
What’s left?
Could an AI agent extract your offer, pricing, and trust signals?
Or would it shrug and move on?
If it’s the latter, you’ve found your first quick win.
Next week in Beyond the Screen, we’ll take this further: Generative UX/UI basics — what it is, where it’s already showing up, and why it’s about to reshape the way we think about interfaces.
Until then, I’d love to hear your thoughts on this:
When you strip your site to text-only, what’s missing?
Hit reply and tell me.
Fiona