# See what AI agents can do in your store, then fix what stops them.

> An automatic site check, key shopping tasks tried by named AI agents in a read-only browser, and a ranked list of fixes for your online store.

- Canonical URL: https://aiscovery.io/product
- Markdown version: https://aiscovery.io/product.md
- Last updated: 2026-10-21

Today: an automatic site check, key tasks tried by named AI agents in a read-only browser, and a ranked fix list with evidence.

- [Request a demo](https://aiscovery.io/demo)
- [Take the tour](https://aiscovery.io/product#tour)

## Example workspace

> Interactive demo on the HTML page: a clickable dashboard with 4 views (Overview, Journeys, Action center, Site check) and a guided tour. Its content is written out below.

> Example data in an illustrative dashboard, not a real store. Store: northwind-outdoor.example.

### Guided tour

1. **The verdict first.** One sentence on your store, then the loop: where agents can read, try, finish and what got fixed.
2. **Every task, every agent.** Pick a cell to open the run: each step, its screenshot and the check that decided it.
3. **Fixes, ranked.** Each fix names the change and who should make it. Press Mark fixed and we re-run the same check.
4. **What AI crawlers may read.** Your robots.txt, read the way each AI company's crawler reads it, plus page basics and AI-facing files.

### Overview

Example verdict: Agents find your products, then get stuck on delivery and returns.

| Stage | Example result |
| --- | --- |
| Readable | 10 of 14 site checks pass |
| Tried | 43 test runs by 3 agents |
| Finished | 37 of 43 reviewed runs worked |
| Fixed | 4 verified by a retest |

### Journeys

| Key task | GPT-6 Luna | Gemini | Computer use |
| --- | --- | --- | --- |
| Find a product and its price | Worked | Worked | Worked |
| Check if it is in stock | Worked | Agent failed, not your site | Worked |
| Find the delivery cost | Stuck on your website | Stuck on your website | Stuck on your website |
| Find the returns policy | Worked | Stuck on your website | Worked |

Test run, Worked:

1. Opened the store
2. Followed a category link
3. Read the product page
4. Found the fact in the page text

- Fact found in page text: yes
- Matches structured data: yes
- AI check confirms: yes

Test run, Stuck on your website:

1. Opened the product page
2. Looked for the fact
3. Followed the footer link
4. Stopped: the fact isn't on the page

- Fact found in page text: no
- Linked from the product page: no
- AI check confirms the failure: yes

Test run, Agent failed, not your site:

1. Opened the product page
2. Picked the wrong size
3. Read stock for the wrong size
4. Answered without checking

- Fact was on the page: yes
- Agent read the right element: no
- Counted against the agent, not your site: yes

### Action center

Fix states: Open, Marked fixed, Retesting, Verified fixed.

**Show the delivery cost as text** (High priority, Policy pages, Who: Content)

Agents can't quote a total: the shipping page links a PDF and products say calculated at checkout.

```diff
  <section id="shipping">
-   <a href="/rates.pdf">Shipping rates</a>
+   <p>Standard delivery: 4.95, free over 60.</p>
```

**Let agents load your page files** (High priority, Page files, Who: Developer)

robots.txt blocks /wp-content/plugins/, so product pages render without their scripts.

```diff
  User-agent: *
- Disallow: /wp-content/plugins/
+ Allow: /wp-content/plugins/*.js
```

**Keep pop-ups off policy pages** (Medium priority, Pop-ups, Who: Marketing)

A sign-up pop-up covers the returns policy until someone clicks.

```diff
  popup.rules:
-   show_on: all_pages
+   exclude: [/pages/returns, /pages/shipping]
```

### Site check

AI crawler rules in robots.txt:

| Crawler | Company | Example rule |
| --- | --- | --- |
| GPTBot | OpenAI | Allowed |
| OAI-SearchBot | OpenAI | Allowed |
| ClaudeBot | Anthropic | Blocked |
| PerplexityBot | Perplexity | Allowed |
| Google-Extended | Google | Allowed |
| Applebot-Extended | Apple | Not mentioned |

Page basics:

- Home page reachable, no redirect loop: pass
- Product structured data: pass
- Sitemap listed in robots.txt: pass
- Delivery terms in page text: fail

## From visible content to verified outcomes.

Aiscovery connects three questions that isolated SEO checks and browser scripts leave apart.

1. **Discoverability.** Can an agent reach, interpret and cite the facts needed for the task? Output: Crawl, metadata, structured content and citation evidence.
2. **Task readiness (Available).** Can an agent complete the intended journey and verify the final state? Output: Key task test runs with step screenshots and checked outcomes.
3. **Monitoring (Coming later).** Did a release, content change or agent update alter the result? Output: Weekly re-checks and an email when something changes.

## Control the run and keep what happened.

Each test run gets an explicit scope, approved pages and a stop point before the agent starts.

### How a test run works

The same four moves every time, so a result means the same thing next week.

1. **Define the scenario.** Goal, prerequisites, allowed actions and final-state assertions become a versioned test contract.
2. **Configure the environment.** The website, allowed pages, locale and safety rules travel with the run.
3. **Run under control.** The agent reads public pages, follows links and uses site search and filters. It stops before login, cart, checkout, payment or any form that sends data.
4. **Verify and preserve.** Literal checks decide the result. An AI check can only confirm it or flag it for review, and step screenshots keep it reviewable.

> Interactive demo on the HTML page: the four steps play out on one example test run.

> Example: Goal: Find white leather sneakers and what delivery costs. Allowed: read pages, follow links, site search, filters. Stops at: login, cart, checkout, forms that send data. Asserts: price found in page text, delivery cost found in page text.

| Scope | Example value |
| --- | --- |
| Website | northwind-outdoor.example |
| Pages | /, /products/\*, /pages/\* |
| Locale | en-US |
| Speed | 1 page per second |
| robots.txt | followed |

Example pages visited: Home, Search, Product, Shipping.

- Price in page text: 75.00: yes
- Delivery cost in page text: no
- AI check confirms the failure: yes

Example result: Stuck on your website.

## What goes in, what comes out

You bring the store and the tasks that matter. You get results you can act on and prove.

### What goes in

- Website scope and allowed domains
- Critical customer workflows
- The named AI agent for each run
- Live or staging website and its safety rules

### What comes out

- Task result and completion timing
- Step log and step screenshots, with sensitive fields masked
- Fixes linked to the step and check that failed
- Prioritized remediation backlog

## A failed run is not automatically a website defect.

Every failure is assigned to the layer the evidence supports: your website, the agent or our side.

- **Website:** Rendering, controls, content or session behavior prevented completion.
- **Agent:** The agent failed although the page showed what the task needed.
- **Policy:** The requested action was outside the agent or audit authorization boundary.
- **Infrastructure:** Our browser, the network or an outside service failed, so the result is inconclusive and never counts against your site.

> Interactive demo on the HTML page: pick a failure to see which layer it belongs to.

| Example failure | Layer | Why |
| --- | --- | --- |
| The delivery cost only appears at checkout | Website | The page didn't contain the fact. That's a fix on your site. |
| The agent read stock for the wrong size | Agent | The page showed it correctly. The agent made the mistake, so it never counts against you. |
| The task needed the agent to log in | Policy | Logging in is outside what our agents may do on a live store, so the run stops by design. |
| Our browser timed out on a network error | Infrastructure | The failure is ours. The result is inconclusive and never counts against your site. |

## One task, several agents.

Coverage is reported by channel so missing API or MCP support never hides browser readiness.

- **Search and citation:** Public facts, crawl surfaces, structured data and source consistency.
- **Browser and UI:** Rendered pages, links, site search and filters, timing and layout shifts.
- **API, MCP and WebMCP (Coming later):** Declared API, MCP and WebMCP surfaces are noted when we detect them. Running them is coming later.

Aiscovery Scout, our own agent, reads each key task the way different agents see it. It points at the traps, the agents each one catches and who should fix it. [Meet Aiscovery Scout](https://aiscovery.io/product/scout)

## Capability status stays explicit.

Only what runs today is marked Included now. Everything else is marked Coming later until it runs.

| Capability | Status |
| --- | --- |
| Automatic site check of public pages | Included now |
| Key tasks tried by named AI agents | Available |
| Comparing several AI agents | Available |
| Tests of logged-in pages | Coming later |
| API, OpenAPI and MCP inventory | Coming later |
| Weekly checks and alerts | Coming later |

## Production safety is part of the product model.

Our checker never logs in, adds to cart or pays. AI agent task runs stop at the same points and before any form that sends data.

[Read the methodology](https://aiscovery.io/methodology)

## See the public audit before it opens.

Review the planned no-signup result and the boundaries of a public site check.

[Open audit preview](https://aiscovery.io/product/demo-audit)

## Product questions

**Q: Is this a replacement for SEO, accessibility or QA tools?**

A: Partly. We check what AI crawlers can read, and named AI agents try your key shopping tasks on public pages.

**Q: Does every website need an API, MCP or WebMCP surface?**

A: No. When a page declares one, we note it as detected. We don't run these tools yet, and a missing one never lowers your result.

**Q: Is monitoring available today?**

A: Weekly checks and alerts are coming later and will be included in every plan when released.

## See it on your own store.

Leave three details. We run the checks on your store first, then walk you through the result.

[Request a demo](https://aiscovery.io/demo)

Have a real workflow to test now? [Apply as a launch partner](https://aiscovery.io/partners)

---

This is the Markdown version of the page for AI agents. The HTML page at the canonical URL is authoritative. Site index: https://aiscovery.io/llms.txt