Can an AI agent sign up, buy or book on your site?
The Proofsource agent test sends a real browser agent to do what a buyer's AI agent will do: search your buyer question, pick a shortlist, open your site and try to sign up, add to cart or book. It stops at the last step and never pays. You see each step, a screenshot, who it named instead, and where it got stuck.
Can an AI agent find you, reach your pricing, sign-up, checkout or booking page, and get to the last step?
AI agent site test, in short
The facts, ready to quote.
- What it tests
- Whether an AI agent can find your brand, reach your pricing and get to your sign-up form, pay step or booking page.
- Tasks
- Sign up, buy or book. You pick one per test.
- Safety
- The agent never pays, books, signs up or sends a form. It stops at the last step.
- What you get
- A step strip, screenshots, the shortlist it picked, the outcome and who it named instead of you.
- Choices
- Browsing country and which browser agent drives the run.
- Offer check
- Product, pricing and cart pages are read without scripts and checked for price, plan, stock, buy button, shipping and returns.
- Cost
- Charged in credits for what the run really used, never above the cap shown before you start.
- History
- Every earlier test stays one click away.
What is an AI agent site test?
Buyers are starting to hand whole tasks to AI agents: find me a meeting notes tool and sign me up, book a table, order this part. These agents open real websites in a real browser, read the page, click buttons and fill forms. If a cookie wall, a bot check, a hidden price or a confusing sign-up flow stops them, the agent moves on to the next brand on its shortlist. An AI agent site test, sometimes called an agentic commerce readiness test, runs that journey on purpose so you see what the agent sees.
Proofsource runs the test with a computer-use browser agent. It starts the way a buyer's agent would: it searches your buyer question, reads the results and picks a shortlist. Then it opens your site and follows the task you chose, sign up, buy or book, until the last step. It stops at the sign-up form, the pay step or the booking page and never fills one in.
Each test runs more than once, because one run can be luck. You get the share of runs that reached the last step, which page a bot check turned the agent away on, how competitors' sites treated the same agent, and a screenshot of every step. That turns a vague worry about agentic commerce into a short list of things to fix.
How it works
A browser agent tries to find you, reach your pricing and get to your sign-up, cart or booking page, then shows each step.
Pick the task
Choose sign up, buy or book, the browsing country, and which browser agent drives the run.
See the credits first
Before anything runs, Proofsource shows the most the test can cost in credits. You confirm or cancel.
The agent searches and shortlists
It searches your buyer question, reads what it finds and picks a shortlist, just like a buyer's agent.
It tries the task on your site
It opens your site, looks for pricing, then follows the path to the sign-up form, cart or booking page.
It stops before anything real
At the pay step, booking confirmation or form submit, it stops. Nothing is bought, booked or sent.
Read the outcome
See every step with a screenshot, how many runs reached the end, and who the agent named instead of you.
See it in the product
Screens from the Proofsource app, drawn with sample data for an example brand.
| Step | Result |
|---|---|
| Find Driffle | Named, 2nd on shortlist |
| Reach pricing | Yes, from the top menu |
| Reach sign-up | Found a sign-up form an agent can reach |
| Last step | Stopped before sending the form |
Run the agent test with the standard browser agent? This run costs up to 60 credits. Each run is charged at what it really cost, never above the cap.
The agent never pays, books or signs up.
| Page | Kind | What an agent cannot see |
|---|---|---|
| driffle.ai/pricing | Pricing | Nothing missing |
| driffle.ai/plans/team | Pricing | Plan name loads only with scripts |
| driffle.ai/checkout | Cart | Sign-in needed before totals |
| driffle.ai/teams | Product | No buy button in first HTML |
Every capability, live today
9 shipped capabilities behind AI agent site test, each in the product now.
Agent test
A browser agent tries to find you, reach pricing, sign up, add to cart or book, and shows whether it got to the last step. It never submits anything.
Task picker
Choose the job the agent tries on your site: sign up, buy or book.
Country and agent picker
Pick the country the agent browses from and which browser agent drives the run.
Credit cap shown before a run
Before a run starts you see the most it can cost in credits and confirm it. Each run is charged at what it really used.
Run table and step strip
Each run shows its steps, its outcome and who was named, and stops at the pay step or the booking page.
Step screenshots
A screenshot of every step the agent took, so you see exactly where it stopped.
Test history
Earlier tests stay one click away with the credits each one used, so you can compare before and after a fix.
Agent test endpoints
Start a test and read its results through the Proofsource API.
Readable offer and agent-ready checkout
Reads your product, pricing and cart pages the way a shopping agent does, before any script runs, and lists what is missing: price, plan, stock, buy button, shipping, returns, or a sign-in wall before totals.
Who it is for
Make sign-up agent-proof
See whether an agent asked to find and sign up for a tool like yours gets to your sign-up form, or gives up and picks a competitor.
Test agentic checkout
Run the buy task to check an agent can find the product, add it to the cart and reach the pay step without a bot check stopping it.
Check booking by agent
Run the book task to see whether an agent reaches your booking page from a plain buyer question.
Find the blocking step
Use the step screenshots to find the cookie wall, bot check or hidden button that stops agents, then rerun to confirm the fix.
Why it matters
See what agents see
Screenshots of every step replace guesses about how agents handle your site.
Safe by design
The agent stops at the last step. It never pays, books, signs up or sends a form.
Know who wins instead
When the agent drops you, the test names the competitors it picked.
Prove the fix
Rerun the same task after a change and compare it with the earlier test.
Known cost up front
The credit cap is shown before every run and you pay for what the run really used.
Google AI OverviewsGoogle AI Mode
AI agent site test: Proofsource and other tools
AI visibility tools measure what AI says, and some also look at how AI agents visit and read your site. Here is how they describe this area on their own pages.
| Capability | Proofsource | Profound | Scrunch | Otterly.AI | Peec AI | AthenaHQ | Rankscale |
|---|---|---|---|---|---|---|---|
| A browser agent tries to sign up, buy or book on your site | YesA computer-use agent runs the task and stops at the last step, with a screenshot of every step. | Not listed | Not listed | Not listed | Not listed | Not listed | Not listed |
| Monitoring of AI agents and crawlers visiting your site | YesIn AI traffic and crawler access; the agent test adds an active run. | YesWhich AI bots access content, how often, real-time bot monitoringSource | YesAgent traffic monitoring, included in all plansSource | YesWhich AI crawlers and agents visit your site, from server logsSource | YesBot visit insights via CDN and log providersSource | Not listed | RelatedRead-only Cloudflare connection for AI visits, Pro and aboveSource |
| Serving AI agents a readable version of pages | YesChecks your offer is readable and your checkout is reachable by an agent. | Not listed | YesServes AI agents a machine-readable version of pagesSource | Not listed | Not listed | Not listed | Not listed |
| Shopping and product visibility in AI answers | YesThe agent test and offer check cover the cart path itself. | YesTracks product visibility in ChatGPT Shopping down to SKU levelSource | Not listed | YesShows whether products appear in shopping cardsSource | YesProduct visibility, win rate, ratings, competitors, source URLsSource | YesAI shopping visibility and product appearance rates versus competitorsSource | YesShopping items and sponsored placements; buybox wins; ChatGPT, AI Mode, CopilotSource |
| AI-readiness check of key pages | YesReads your pricing, product and cart pages the way a shopping agent does and lists what is missing | RelatedOpens up content AI crawlers cannot readSource | YesDeep audits flag access controls, JavaScript rendering issues, content gapsSource | YesStructured data and AI-readiness review with fixesSource | YesSitemap, llms.txt and robots.txt crawlability historySource | RelatedDetects pages AI crawlers cannot access; not a full auditSource | YesAI readiness score; 90+ checks; fixes; 10 to 200 audits per month by planSource |
Where each of them shines
Deep agent analytics on which AI bots visit your content, plus SKU-level ChatGPT Shopping tracking. Compare in full
An Agent Experience Platform that serves AI agents a machine-readable version of your pages at the CDN level. Compare in full
Agent analytics from server logs with simple log connections, including a WordPress plugin. Compare in full
AI shopping analytics with win rate, ratings and competitor views, plus crawler analytics from CDN logs. Compare in full
Shopping visibility and commerce optimization that improves product content for AI recommendations.
Rankscale tracks 17+ engines on every plan and analyzes shopping items and sponsored placements.
Questions about AI agent site test
What does the AI agent site test actually do?
It sends a computer-use browser agent through the same journey a buyer's AI agent would take. The agent searches your buyer question, picks a shortlist, opens your site and tries to sign up, buy or book. It records each step with a screenshot and stops at the last step.
Will the agent buy something, book a slot or create an account?
No. The agent never pays, never confirms a booking and never fills in or sends a form. It stops at the pay step, the booking page or the sign-up form, and the test records that it got there.
Why does the test run more than once?
One good run can be luck. Agents choose different paths and pages load differently, so the test runs several times and tells you how many runs reached the last step.
What is agentic commerce readiness?
It is how easily an AI agent acting for a buyer can find your offer, understand it and complete the task on your site. The agent test measures it directly: found you, reached pricing, reached the sign-up, cart or booking page.
What usually stops an AI agent on a website?
Common blockers are bot checks that turn automated browsers away, cookie walls, prices hidden behind a contact form, and sign-up or checkout buttons the agent cannot recognize. The step screenshots show which one it hit.
What does the readable offer check look at?
It reads your product, pricing and cart pages as a shopping agent does, from the first HTML before any script runs. It lists what an agent cannot see there: the price, the plan, stock, a buy button, shipping, returns, or a sign-in wall in front of the totals.
Does it test competitors' sites too?
Yes. The test notes how competitors' sites treated the same agent, for example whether a bot check turned it away or let it in, and who the agent named instead of you.
How much does an agent test cost?
Agent tests are charged in credits. Before a run starts you see the most it can cost, and each run is charged at what it really used, never above that cap.
Is this the same as AI crawler tracking?
No. Crawler tracking shows which AI bots visited your pages. The agent test actively tries a task on your site and reports whether an agent can finish it.
Can I choose the country and the agent?
Yes. You pick the browsing country and which browser agent drives the run, so you can test the experience buyers in each market get.
Keep exploring
Try it on your own brand.
Proofsource is live: scan your brand and see AI agent site test at work with your own answers, competitors and sources.
Google AI OverviewsGoogle AI Mode