Can an AI agent read your pricing page and submit your demo form? A dev checklist for agent-ready sites
TL;DR: Browser agents act for one person in real time: comparing suppliers, reading pricing, booking demos. Whether an agent can use your site comes down to four things: your CDN lets it in, your prices are in the HTML,
TL;DR: Browser agents act for one person in real time: comparing suppliers, reading pricing, booking demos. Whether an agent can use your site comes down to four things: your CDN lets it in, your prices are in the HTML, your first form step has no CAPTCHA, and your facts match third-party listings. You can test all four in an afternoon.
How is this different from crawlers?
In my last post I covered sorting AI crawlers in robots.txt: training bots, search bots and user-triggered fetchers. This post covers what happens after a bot gets through the door.
A browser agent is a different kind of visitor. It doesn't index your site or collect text for a future model. It acts for one person, right now: "compare three suppliers and book a demo with the cheapest one that supports SSO."
This isn't hypothetical anymore. OpenAI's dots, announced in September 2026, run their own cloud browser, connect to over 4,000 apps, and can sign in to supported sites. Cloudflare now treats this as its own traffic class: its July 2026 changelog splits AI traffic into Search, Agent and Training, each with its own controls.
If the agent can't read your price or submit your form, it moves on to a competitor whose site it can use.
1. Is your CDN quietly blocking agents?
Even with a perfect robots.txt, a CDN rule can stop an agent before it sees anything.
One detail to watch: per Cloudflare's changelog, from 15 September 2026, new domains block Training and Agent traffic by default on pages that show ads. If you launched a microsite or moved a domain recently, check it.
Decide per page group, not per site:
| Page group | Agent traffic |
|---|---|
| Pricing, plans, plan comparison | ✅ Allow |
| Product and feature pages | ✅ Allow |
| Contact / book a demo | ✅ Allow |
| Admin, account, checkout | ❌ Keep protected |
Also check scraper-focused WAF rules. They're often written broadly enough to catch agents too.
2. Is your price in the raw HTML?
Agents work best with what's in the page source. If your price only appears after a client-side fetch, a tab click, or inside an image, assume the agent might miss it.
A quick check:
# Does the price appear in the server-rendered HTML?
curl -s https://example.com/pricing | grep -iE "£|\\$|€|per month|/mo"
If that returns nothing but you can see prices in the browser, they're being rendered client-side.
Put these in server-rendered text for each plan:
<section aria-labelledby="plan-growth">
<h2 id="plan-growth">Growth</h2>
<p><strong>$499 per month</strong>, billed annually (USD)</p>
<ul>
<li>Up to 10 users</li>
<li>50,000 API calls per month</li>
<li>Minimum term: 12 months</li>
</ul>
</section>
Plan name, price, billing period, currency, main limits and minimum term all belong in that text. If you don't publish prices, publish the pricing structure and what each plan includes. "Contact us" on its own gives an agent nothing to compare.
What about structured data? Schema markup may help, but don't rely on it, and it must match the visible text exactly. If your JSON-LD says $449 and the page says $499, you've created a conflict the agent has to resolve, and it may not resolve it in your favour.
3. Can an agent complete your form?
Most demo forms are built for a human with a mouse. Agents need something more predictable, and it overlaps almost entirely with good accessibility practice:
<form action="/demo" method="post">
<label for="name">Full name</label>
<input id="name" name="name" type="text" autocomplete="name" required>
<label for="email">Work email</label>
<input id="email" name="email" type="email" autocomplete="email" required>
<label for="phone">Phone</label>
<input id="phone" name="phone" type="tel" autocomplete="tel">
<label for="company">Company</label>
<input id="company" name="company" type="text" autocomplete="organization">
<button type="submit">Request a demo</button>
</form>
<!-- On success: confirm in page text, not just a toast or modal -->
<p role="status">Thanks. Your demo request has been received. We'll email you within one working day.</p>
The rules:
-
A visible
<label>for every field. Placeholder text is not a label. -
Standard input types, especially
type="email"andtype="tel". - No mouse-only controls: no drag sliders, custom dropdowns without keyboard support, or "click the image" pickers.
- Confirm success in page text. A pop-up that disappears after two seconds is easy for an agent to miss.
What about spam?
You don't have to choose between agents and spam protection. Use invisible checks, rate limits, and challenges on later steps, not a CAPTCHA on the very first submission. Booking a demo is a reasonable thing for an agent to do for someone. Accepting terms or paying is where it's sensible to make the human step in.
4. Do your facts match everywhere else?
Agents cross-check. If your pricing page says "14-day trial" but a review site says "30-day trial" and a directory lists last year's plan names, the agent has to guess which to believe.
Keep a short list of your "canonical facts" (prices, plan names, minimum term, trial length) and check them against the directories, review sites and comparison pages that list you.
How do you test all this in one afternoon?
Run these in order, and fix each blocker before moving on:
- CDN: read the Search / Agent / Training rules for each page group.
-
Page source:
curlyour pricing page and confirm price, limits and minimum term are in the raw HTML. - Agent task: ask a browser agent to compare suppliers in your category and book a demo with you.
- Form result: check that the submission actually arrived, and look at exactly what the agent typed into each field.
- Listings: compare your canonical facts against third-party listings.
Two caveats. Nobody has yet published solid data on how many buyers use browser agents. And agent visits in your CDN or server logs are a floor, since some agents don't identify themselves or run your analytics scripts.
Where can you go deeper?
This post is the developer's version of a longer piece from the Sigzen AI team: Can an AI agent read your prices and complete your forms?. It covers the commercial side too: which pages to prioritise, what to let agents do, and how to test across different markets.
If you want to see how AI assistants describe your product before an agent ever visits your site, there's a free AI visibility score. It checks whether ChatGPT, Gemini, Copilot and Google's AI features name your domain and which pages they cite.
Have you tried pointing a browser agent at your own signup or demo flow? What broke first? Tell me in the comments.
Written by the Sigzen AI editorial team. Sigzen AI helps B2B brands show up accurately in AI answers and builds AI agents for operations teams. Previous post: Your robots.txt is probably blocking the wrong AI bots.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.