Aiscovery Scout · our agent

Our agent looks at your store the way each AI agent does.

Aiscovery Scout reads each key task as a screenshot, as page text, without scripts and through robots.txt, then shows you the traps before an agent runs into them.

northwind-outdoor.example/products/court-low-sneaker
What people seeScreenshot

Northwind

  • New
  • Men
  • Women
  • Sale

Search for:Search

First screenshot ends here

Court Low Sneaker

75.00 Add to cart

Example data, not a real store.

In your dashboard: Possible traps for AI agentsAvailable

The traps Scout looks for

Each one comes with the evidence Scout saw, the agents it affects and who should fix it.

  • Page textPage without scripts

    The fact isn't on the page

    The price, stock or delivery cost a key task looks for is missing from the page, or has changed.

    AffectsAll agentsFix on your websiteUpdate the key task
  • Page without scripts

    No path by links

    The page an agent needs isn't linked from where the task starts, or a link on the way is broken.

    AffectsAll agentsFix on your website
  • robots.txt

    Page files blocked

    robots.txt blocks the styles and scripts a page needs, so a browser that follows it draws the page broken.

    AffectsComputer use agentsFix on your website
  • Screenshot

    The screenshot shows less

    One element fills the first screen, or every page looks the same, so an agent working from screenshots has nothing useful to click.

    AffectsComputer use agentsFix on your websiteOn our side
  • Task rulesrobots.txt

    Rules that close the way in

    Search is the main way to find the product, but the task rules don't let agents type, or robots.txt closes search.

    AffectsAgents that type in searchChange the task rulesFix on your website
  • Task rules

    Not enough time

    An agent usually needs longer than the key task allows, so it would stop halfway through.

    AffectsAgents that need longerChange the task rules
  • Still being calibrated

    Pop-ups over the price, prices only in an image, buttons without names, content inside frames, bot walls and duplicate buttons. We show them once real agent runs back them up.

    Coming later

How Scout works

The same four moves on every key task, so a trap means the same thing next time.

  1. Step 1
    Read the key-task pathScout takes a key task and the pages it should pass through, inside the rules the store owner approved.
  2. Step 2
    Look through each agent's eyesOn every page it takes a screenshot, reads the page text, loads the page without scripts and checks robots.txt and the task rules.ScreenshotPage textPage without scriptsrobots.txt
  3. Step 3
    Evidence, owner and fixEach trap shows what Scout saw, which agents it affects and who should act: your website, the key task, the task rules or us.
  4. Step 4
    Real agent runs confirm or refuteWhen agents try the same task, their runs show whether Scout was right: a real agent run failed here too, or agents got past it.A real agent run failed here tooAgents got past it

Two jobs on the same store.

A shopper's agent and Scout look at your pages for different reasons. You need both.

A shopper's agenttries the task once.

  • Reads the page one way: its own
  • Follows one path to an answer
  • Shows where it got stuck, after it happened

Aiscovery Scoutchecks every way agents read the page.

ScreenshotPage textPage without scriptsrobots.txtTask rules
  • Sees each page the way different agents do
  • Walks the key-task path page by page
  • Points at the trap, who it catches and who fixes it, before a run

Scout doesn't replace real agent runs. It shows where to look, and the runs show whether it was right.

Read-only, inside the rules you approved.

Scout reads public pages on the key tasks the store owner approved. It reads robots.txt first, goes one page per second and stops before login, cart, checkout and payment. It never logs in, fills in forms or pays.

If a bot wall or a firewall stops it, Scout says Couldn't check and never tries to get around it. Your agent test runs come first: Scout waits while one is running on your site.

Scout stops before

  • Login
  • Cart
  • Checkout
  • Payment
  • Forms that send data
  • Read-only
  • Only what the owner approved
  • robots.txt first
  • One page per second
  • Never logs in or pays

What runs today

Scout runs for launch partners now. Everything else is labelled.

  • Scout on your key tasks, for launch partnersAvailable
  • Traps with evidence, affected agents, owner and fixAvailable
  • More traps: pop-ups, prices in images, unnamed buttonsComing later
  • Weekly Scout with alertsComing later
  • Choosing the AI model that double-checks screenshotsComing later

Questions about Scout

Is Scout a score for my store?

No. Scout gives no single score. It shows concrete traps, each with evidence, the agents it affects and who should fix it. Clear means Scout found no trap in that lens, not that agents will succeed.

Does Scout replace real AI agent test runs?

No. Real agents try your key tasks and show what happens. Scout looks first, so you can fix the obvious traps before you spend a test run on them, and the runs show whether Scout was right.

Does a Scout check use up my test runs?

No. Scout runs separately and doesn't count against your plan's monthly test runs.

What if Scout can't see a page?

If a bot wall, a firewall or robots.txt stops it, Scout says Couldn't check for that page. It never gets around protection, and it never counts that against your site.

See the traps on your own store.

Leave three details. We run Scout on your key tasks first, then walk you through what it found.

Have a real workflow to test now? Apply as a launch partner

We only use this to arrange your demo. Privacy notice