The Add to cart button has no name
00:14 UTCGive button.cart-add a text or a label on every product page.
Can AI agents use your website?
For a full day, every day, AI models work through your site without stopping on a fresh machine and report everything that kept an AI agent from discovering, understanding, navigating, or using it, with what they saw and what to change.
shop.example.comFull-day review · 00:00 to 24:00 UTC
Review done- No llms.txt at the site rootLow
- /llms.txt answers 404.
- Prices appear only after scripts runHigh
- The HTML of /pricing holds no price, only the script that loads it.
- The Add to cart button has no nameHigh
- button.cart-add has no text, label, or alt.
- The cart sends every visitor to sign inMedium
- /cart answers 302 to /login.
2out of 4Questions without a high risk
0out of 4Questions without a high risk
What an AI agent meets on your site
One report on shop.example.com, quoting what the agent met for each question it works through: can it discover, understand, navigate, and use the site.
Discover
Whether an agent may read the site and can find its pages.
/llms.txtcurl
$ curl -sI shop.example.com/llms.txtHTTP/2 404content-type: text/html
Publish an llms.txt at the site root that names the key pages.
Understand
Whether the facts on a page are there for a machine, not only for a person looking at it.
/pricingHTML the server sent
<main id="app"> <h1>Pricing</h1> <div class="price" data-plan="pro"></div> <script src="/assets/prices.js" defer></script></main>
Put the price in the HTML the server sends, or in structured data.
Navigate
Whether an agent can find its way through the site by the names of its links and controls.
/product/42HTML the server sent
<div class="product" data-sku="42"> <h1>Trail Runner 2</h1> <button class="cart-add"> <svg aria-hidden="true">…</svg> </button>
Give button.cart-add a text or a label on every product page.
Use
Whether an agent can reach what a visitor comes for, such as a price, a product, or the cart.
/cartcurl
$ curl -sI shop.example.com/cartHTTP/2 302location: /login?next=/cart
Let a visitor see the cart and its prices before signing in.
Every risk this month
What agents met on shop.example.com this month, with how many of the pages they read each risk concerned.
Pages
Discover
Understand
Use
Product pages missing from the sitemap4
robots.txt turns some AI agents away1
No llms.txt at the site root1
Page titles that repeat on every page9
Prices appear only after scripts run6
No structured data for products5
Product facts only in images3
Menu links with no text8
The Add to cart button has no name5
Search results only after scripts run1
The cart sends every visitor to sign in1
Order form fields without a label1
A sample review
One day’s report of shop.example.com, the way a subscribed team reads it.
4Risks in today’s report2 high, 1 medium, and 1 low
7
Mon7
Tue6
Wed6
Thu5
Fri5
Sat4
Sun14Pages readof the 120 links the agent saw
12
Mon13
Tue12
Wed14
Thu13
Fri15
Sat14
SunPages read without a risk
43percent of pages read
What to change, not only what is wrong
For each risk, the agent says what the site could change so agents get through.
Today
Prices appear only after scripts run
00:14 UTCPut the price in the HTML the server sends, or in structured data.
The cart sends every visitor to sign in
00:14 UTCLet a visitor see the cart and its prices before signing in.
No llms.txt at the site root
00:14 UTCPublish an llms.txt at the site root that names the key pages.
Today
The Add to cart button has no name
00:14 UTCGive button.cart-add a text or a label on every product page.
Prices appear only after scripts run
00:14 UTCPut the price in the HTML the server sends, or in structured data.
Today
The cart sends every visitor to sign in
00:14 UTCLet a visitor see the cart and its prices before signing in.
Today
No llms.txt at the site root
00:14 UTCPublish an llms.txt at the site root that names the key pages.
Risks by question
4
Today
1A week ago
2Where the agent went
120
Links the agent saw
120Pages read
14Pages with a risk
8Pages with a high risk
6Where the risks are
Risks
By page
By severity
/product/422
/product/72
/product/92
/product/122
/product/152
High2
Medium1
Low1
| Risk | What to change | |||
|---|---|---|---|---|
| The Add to cart button has no name | Navigate | High | 5 | Give the button a name |
| Prices appear only after scripts run | Understand | High | 6 | Put prices in the HTML |
| The cart sends every visitor to sign in | Use | Medium | 1 | Show the cart before sign-in |
| No llms.txt at the site root | Discover | Low | 1 | Publish an llms.txt |
Risks found
94
Risks over two weeks
94Pages the agent read
14From a verified site to a review that runs all day
Three steps, and then a full day of review every UTC day, with no one starting it.
Set up once
Prove you control the site, accept the review agreement, and the first review runs that UTC day.
Ready for the daily review
- Verify the site
- Accept the review agreement
- Run the first review
A full day of review, every UTC day
At 00:00 UTC Riskcovery starts a fresh Linux machine, and the models on it work on the site until the day ends, keeping the report current as they find more. Then the machine is deleted.
Verify the site
A DNS TXT record, a meta tag, or a verification file. Riskcovery checks the proof again before every run.
$ dig +short TXT _riskcovery.shop.example.com "riskcovery-site-verification=7f3a9c…"
Accept the review agreement
An owner or admin accepts it, and can revoke it at any time.
- Hosts
- shop.example.com
- Schedule
- One run every UTC day
- The agent
- Only reads
- Revoke
- Any time
What one run does
The agent keeps working through the day, and every risk names the pages it concerns and quotes what the agent saw there.
Reviewing shop.example.com all day
Checked the DNS TXT proof
VerifiedStarted a fresh Linux machine
This run onlyRead the home page, robots.txt, and the sitemap
curlFollowed links to the product pages and the cart
14 pagesQuoted button.cart-add on /product/42
HighKept the report current until the day ended, then deleted the machine
4 risks
Risks found each day this week
40
Discover
Understand
Navigate
Use
MonTueWedThuFriSatSunFewerMore
Risks by day
4
Read the review wherever you are
The website, the CLI, and the MCP server ask the same server for the same reports.
Inside your agent
An MCP server signed in through OAuth, so your own agent can read its reviews.
What did today’s review of shop.example.com find?
- shop.example.com · today, 00:00 UTC
- 4 risks · review done2 high
Today’s review found 4 risks, 2 of them high: the Add to cart button has no name on the product pages, and prices appear only after scripts run. The report says what to change for each.
From the terminal
List sites, request today’s run, and print a report as JSON or Markdown.
$ npx riskcovery review shop.example.com 4 risks · 2 high High The Add to cart button has no name High Prices appear only after scripts run Medium The cart sends every visitor to sign in Low No llms.txt at the site root
In your tracker
Export a report as Markdown or JSON and paste it where your team tracks work.
## High - The Add to cart button has no name (5 pages) Give button.cart-add a text or a label. - Prices appear only after scripts run (6 pages) Put the price in the HTML the server sends.
On the web
Every earlier run of your site, with share links for a full report.
| Run | |||
|---|---|---|---|
| Today, 00:00 UTC | 4 | 2 | 14 |
| Yesterday, 00:00 UTC | 5 | 2 | 15 |
| 2 days ago, 00:00 UTC | 5 | 3 | 13 |
An agent you can trust on your own site
Riskcovery sends an AI agent only where it was invited, and publishes how it reads a site.
- Reads only sites whose owners verified control and accepted a review agreement
- Tells the agent to only read: never to sign up, log in, buy, post, or submit a form
- Points the agent at the site under review and at no other site
- Runs each review on a fresh Linux machine and deletes it when the run ends
What today’s agent read17 / 17
- Pages of the site14
- robots.txt, the sitemap, and llms.txt3
- Runs a dayResets 00:00 UTC1 / 1
- Machines a runDeleted when the run ends1 / 1
Questions
What does Riskcovery review?
Whether AI agents can discover, understand, navigate, and use your website. Every UTC day, AI models work on the site for the whole day without stopping, on a fresh machine, and report what got in an agent’s way, with what they saw and what the site could change.
How is this different from AI search visibility tools?
Search visibility tools report what answer engines say about a brand. Riskcovery sends an agent to the site itself and reports what it met there, with evidence a developer can act on.
How often does a review run?
One run every UTC day for your site once its agreement is accepted, while your team holds a subscription. A run starts at 00:00 UTC and its models keep working on the site until the day ends, updating the report as they find more; a site’s first run works for the rest of the day it starts. Every run gives the agent the same goal, on the fast model, or on the premium model on a team’s premium day.
What does a subscription show?
A subscription shows the whole report of each run, with exports and share links.
Can someone review a site they do not own?
No. Riskcovery reads a site only after its owner verified control and accepted the review agreement, and the owner can revoke it at any time.
How do I get access?
Riskcovery is invitation only. Request an invitation with your work email and the site you want reviewed, and staff send one when a place opens.
See what agents cannot do on your site
Request an invitation, verify your site, and read the first findings the same day.