Blog · 7 min read

Your next customer might not be human. Get your website ready.

Dušan Kníže · August 24, 2026

← Back to Blog

Picture a customer who never sees your website. They don't click a banner, don't read your slogan, don't notice the carefully tuned design. Instead, their AI assistant arrives, scans your offer in a second, compares it with three competitors and recommends — or simply places the order. That scenario has stopped being science fiction. In 2026, it's the fastest-growing trend in e-commerce.

It's called agentic commerce — shopping mediated by AI agents. And because it's mostly written about in the context of American e-shops, it's easy to wave off as something distant. That would be a mistake. Let me explain why — and above all, what to practically do about it.

What exactly is happening

The numbers from recent months are steep even by AI-world standards. According to Adobe, traffic to US retail sites coming from AI tools grew 693% year over year during last year's holiday season. Shopify reports that AI-driven traffic to its stores grew eightfold in the first quarter of 2026. And surveys show 73% of consumers already use AI somewhere in their shopping journey — most often for product ideas, review summaries and price comparisons.

Banks and consultancies are racing to estimate where this leads: J.P. Morgan expects that by 2030, up to a quarter of US online sales could flow through AI agents. Take the exact figures with a grain of salt — they're projections. But the direction is unambiguous.

The fundamental shift: which company a customer sees is increasingly decided not by advertising or search rankings, but by whether an AI assistant can read and understand your website. If you're unreadable to agents, you don't exist for their users.

Why this concerns smaller markets too

Smaller European markets lag months, maybe a year, behind the US. But the mechanism is already running here — just in a simpler form. When someone asks ChatGPT or another assistant to "find me a reliable accountant" or "compare solar panel suppliers", the assistant genuinely crawls websites and assembles an answer. Companies whose pages it can read are in that answer. The rest aren't.

This isn't the future — it's happening now. I see it in my own website's analytics, where visitors arrive from ChatGPT and Perplexity. And unlike classic SEO, where rankings take years to win, this discipline is so new that a small company can leapfrog a big one simply by having a technically sound website.

How an AI agent "sees" your website

To understand what to fix, it helps to know how such an agent works. It's usefully mundane: the agent downloads the page's HTML and reads. No eyes, no taste — just text and structure.

Three uncomfortable truths follow:

The agent doesn't see what JavaScript renders. Most AI crawlers don't execute scripts. If your prices, descriptions or contact details only appear after scripts load, the agent sees an empty page. A beautiful animated site built as an app can be invisible to agents — while a competitor's plain static page is perfectly readable.

The agent doesn't read images. A price list embedded as an image, opening hours in a graphic, a phone number in a banner — none of that exists for the agent. What isn't text isn't information.

The agent loves structure. Machine-readable data (schema.org — prices, services, questions and answers) gets read preferentially, because the agent can quote it without guessing. A site with structured data has an edge over an equally good site without it — an edge it never had with classic Google.

What to actually do — in order of importance

1. Verify your site is readable without JavaScript. Disable scripts in your browser's developer tools and load your site. Can you see prices, services, contact details? You're fine. Do you see a blank page? You have a problem bigger than the rest of this list combined.

2. Get prices and offers into text and structured data. Concrete prices in the HTML, and ideally in schema.org markup too. I know plenty of companies "prefer not to list prices" — but an agent that can't find a price simply drops you from the comparison. A published price is the ticket into the game.

3. Let AI crawlers in. These days robots.txt files often block everything "AI", because that was once recommended over model training concerns. But that also blocks the agents shopping on customers' behalf. Decide consciously — blocking search and shopping agents means giving up that traffic.

4. Add llms.txt. A new standard: a text file that summarises for AI tools what your company does, what it offers and at what price. Written once, works forever. I run one on my own site, generated automatically from the price list.

5. Answer the questions people ask their agents. An FAQ section with structured data is a gold mine for agents — exactly the format they assemble answers from. "How much does… cost", "how long does… take", "does it work with…" — the more specific, the better.

If this list reminds you of good SEO, that's no coincidence. Most of the preparation for agents overlaps with what helps a site in classic search engines — I build websites this way from the start, and it works for both worlds at once.

What not to do

Don't panic-buy "AEO packages". The same wave of overpriced services that once grew around SEO is now growing around AI search optimisation. Most of what you need is described above, and it's ordinary craftsmanship, not magic.

Don't fake data. Inflated reviews and invented ratings in structured data are tempting, because "only a machine reads it". But agents increasingly verify claims across multiple sources — and an inconsistency is a worse signal than a missing value. The same rule as with people: trust builds slowly and evaporates fast.

Don't rebuild your site for fashion's sake. If your site serves people well today, it doesn't need a rebuild — it needs machine readability added. That's an order of magnitude smaller job.

An honest conclusion

Shopping AI agents are a marginal phenomenon in smaller markets today — and anyone claiming otherwise is selling you something. But the preparation is cheap, overlaps with good technical practice, and its effects compound: a site readable by agents is also faster, more accessible and better ranked by search engines. This is the rare case where you can prepare for a trend with no risk of the investment being wasted, even if the trend slows down.

And one more thing: the next time someone evaluates your website through an AI assistant, the decision takes seconds and carries no emotion. There's no design to charm them, no salesperson to persuade them. Just the data you prepared for them — or didn't.

Frequently asked questions about shopping AI agents

Shopping mediated by AI agents — the customer states what they need, and their AI assistant browses offers, compares them and recommends or directly orders. In the US it's the fastest-growing e-commerce segment (AI-driven traffic to retail sites grew by hundreds of percent year over year during the holiday season). In smaller markets a simpler form is running: AI assistants answering "find me a supplier" questions based on what they can read from websites.

The fastest test: disable JavaScript in your browser and load your site. What you see is what the agent sees — whatever disappeared doesn't exist for it. The second step is checking robots.txt (are you blocking AI crawlers?) and verifying that prices and key information live in text, not images. I do a complete check of this kind as part of an AI audit.

Usually not. If the site serves people well, it's about adding machine readability: structured data (schema.org), prices in text, llms.txt, allowing AI crawlers, and an FAQ section. That's days of work, not a rebuild. The exception is sites rendered purely by JavaScript — there you need to make sure the content also exists in the HTML.

For AI agents it's a decisive advantage. An agent comparing offers drops any company whose price it can't find — there's nobody to ask. A company with public prices enters a comparison its price-hiding competitors are excluded from. It's the same principle my own price list is built on: concrete amounts, machine-readable.

It depends on what your website is for you. For a content business (media, paid courses), restrictions make sense — the content is the product. For a company that wins customers through its website, blocking AI crawlers is counterproductive: you're blocking a channel customers arrive through. A sensible compromise is to allow search and shopping agents while restricting purely training-focused crawlers.
Want to know how AI agents see your website?

As part of an AI audit I'll go through your site with AI crawlers' eyes — what they read, what they're missing and what specifically to add, ranked by impact. No panic and no overpriced packages.

Book a free 30-minute call
DK
Written by Dušan Kníže
AI developer · Prague, Czech Republic

I build AI solutions for businesses — from chatbots to knowledge systems. I write about what actually works, no buzzwords. More about me →