METHODOLOGY · UPDATED SEPTEMBER 5, 2026

What we test.
What it proves.

Three kinds of evidence answer different questions. We name the method on each report and keep untested steps visible.

1. The free HTTP check

We fetch public pages, robots.txt and available structured data from our server. We compare a browser-like request with four published user-agent strings: ChatGPT-User, Perplexity-User, Amazonbot and Claude-User. These are request identities, not four consumer agents.

A site URL checks the home page and, on stores, discovers a product page; a specific URL checks that page. The report shows the selected URL and timestamp. Requests have an eight-second timeout, a two-megabyte response cap and a 40-second overall budget. Results may be cached for ten minutes.

On stores we inspect Product / ProductGroup JSON-LD for price, stock, identifiers, shipping, returns and ratings, and fetch the cart and checkout paths. Missing markup is a data signal. It does not establish that the same information is absent from the visible page or a feed. On other sites these rows show as untested.

Not tested: JavaScript interactions, real agent network identities, menus and cookie banners, form entry, cart contents, bookings, payment completion, recommendation ranking or revenue. A loaded CAPTCHA script does not establish that an action is blocked. Network errors and timeouts are reported as unassessed.

2. Automated browser observations

A scripted browser can open a page, follow a menu, fill a form or inspect a cart. This is a reproducible automation test, not a consumer AI agent. A selector failure can be a harness limitation; we review it before attributing a problem to the site.

Browser observations are not presented as proof of real-agent completion or lost business. Public examples identify their method. Illustrative deliverable formats are labeled separately from measured results.

3. The real-agent run

We agree on your site, the five tasks in your words, and three available consumer agents. Each agent attempts each task, recorded. Agents are confirmed before payment; no platform is silently substituted.

Each attempt records the prompt, agent and date, the furthest step reached, what was on screen when it stopped, and any obstruction. We distinguish observed behavior, a likely cause and a confirmed cause.

We stop before submitting a payment, accepting an agreement or sending anything on your behalf. Any login, approval or required human takeover is recorded as a handoff, not a defect. We do not bypass security challenges, alter site settings or need admin access.

The deliverable contains all outcomes, including passes, prioritized developer tickets with reproduction steps, and acceptance checks. One re-run of the same attempts is included when requested within 14 days of delivery. Broader testing is quoted separately.

Timing: scope and availability are confirmed before payment; delivery is within two business days after confirmation and payment. If we cannot run the agreed test, we decline it before charging. Passing outcomes do not trigger a refund: the product is documented evidence, not a promised failure or revenue lift.

4. Monitoring

The agreed tasks are re-run on a schedule and whenever a listed agent ships a notable update. You get the dated result of each run, an alert when a step that passed starts failing, and the same runs on sites you name for comparison. Comparison sites are tested with public prompts on public pages only, and are never contacted.

5. The fix package

A fixed scope agreed by email from a run’s findings: cookie banner and overlay behavior for agents, guest paths for forms and checkout, structured data, agent-readable versions of key pages, a basic MCP endpoint. Changes are delivered as tickets or pull requests for your team to review; nothing ships to your production without your sign-off. The re-run that shows the result is included.

How to interpret the result

Fifteen attempts are a small diagnostic sample. They do not estimate a population completion rate. A failure in one dated run does not show that all agents or all people fail. A pass is not a certification. No score here predicts revenue, recommendation placement or agent traffic.

We recommend verifying a finding in the relevant environment before changing access controls, forms, checkout settings or data feeds.

Data handling

The free check needs only a public site or page URL. Do not include access tokens, private preview links, passwords or customer details. Results are held in an in-memory cache for up to ten minutes and displayed in your browser; they are not automatically added to a public index.

Copy and download controls create a report on your device. If you request a run by email, the site, pages, email, role, optional notes and check summary are used to scope and deliver that request, not to subscribe you to marketing.

We log basic funnel events, such as a completed check or an opened request, with campaign labels. The event payload does not include your email, site URL, cookies or a browser fingerprint. The hosting provider may retain standard server request logs. Contact hello@usecreel.com to ask about or delete your submitted request.

Source context

Shopify’s product-page scanner describes its output as informational product-page signals. OpenAI’s product discovery update describes Shopify Catalog as a source of merchant product data. That is why missing page markup alone is not proof that an agent cannot find or use a page.