Watch Owl: website QA that tests every page in a real browser
Give Watch Owl your website address. It opens up to 500 of your pages in a real Chromium browser, uses them the way a visitor would, and runs 54 checks on the pages and on the site around them. You get a fix list, worst problem first.
What happens after you press start
A crawl is six passes over your site. The first five happen on each page in turn; the sixth looks at the site as a whole.
- It opens your homepage in headless Chromium, the engine inside Chrome, so your JavaScript runs exactly as it does for a visitor. A site that only renders in the browser gets tested properly, not as an empty shell.
- It follows every link to your own pages, up to 500 per crawl. For each page it records the status code, every console message and every network request that failed or timed out.
- It runs the axe-core accessibility rules and checks every image. Missing labels, low contrast and images that never load are all logged against the page they happened on.
- It uses your forms. Text fields get test values, and the form is submitted. A form on anything that looks like checkout or payment is never submitted: it won't press pay, buy now or place order. If a page sends the browser off to an outside sign-in such as Google, it stops touching that page.
- It takes a full-page screenshot and compares it with the same page from your previous crawl. A page whose pixels changed by more than 2% is flagged, which catches layout breaks nobody reported.
- It checks the site around the pages: DNS, SSL, domain expiry, redirects, robots.txt, sitemaps, email records, security headers and server response time. Then it compiles everything into one report.
You can watch findings arrive while the crawl runs, and stop it at any point. Whatever it has found up to then stays in your report.
All 54 checks, grouped as your report shows them
These are the exact checks in a Watch Owl report. Each one comes back as passed, a warning or a failure, with the page and the evidence.
On every page
7Measured in a real browser, page by page.
- Internal links that return an error
- JavaScript console errors and uncaught page errors
- Failed network requests (4xx, 5xx, timeouts)
- Accessibility rules (axe-core)
- Forms that throw an error when used
- Images that fail to load
- Visual change since the last crawl
Domain and security
9Checked once for the whole site.
- DNS resolution
- Domain expiry date
- SSL certificate
- HTTP to HTTPS redirect
- Mixed content
- Security headers
- Exposed .env file
- Staging or dev leaks
- Placeholder content left live
Email, trust and tracking
9The things visitors and inbox providers check before trusting you.
- SPF record
- DKIM record
- DMARC record
- Cookie consent banner
- Form spam protection
- Trust pages (About, Contact, Privacy, Editorial)
- Favicon
- Open Graph tags
- Analytics tracking tag
Crawling and indexing
10Whether search engines can reach and trust your pages.
- robots.txt rules, per bot
- robots.txt blocking CSS or JavaScript
- XML sitemap: URL count and file size
- Sitemap URLs that don't load
- Sitemap index resolution
- Sitemap vs. canonical consistency
- Noindex pages listed in the sitemap
- Canonical target health
- Redirect chains and loops
- URL variant consistency (www, trailing slash)
Titles, content and markup
10Page-level search basics, compared across the whole site.
- Missing title tags
- Duplicate title tags
- Missing meta descriptions
- Duplicate meta descriptions
- Duplicate content without a shared canonical
- Schema types found
- Hreflang tags
- Image alt text coverage
- Image width and height coverage
- Pages set to noindex or nofollow
Mobile, structure and speed
9How the site holds together as one thing.
- Mobile viewport tag
- Intrusive mobile pop-ups
- Mobile vs. desktop content parity
- Orphan pages
- Click depth from the homepage
- Pages with no internal links pointing in
- Pagination sequence gaps
- Faceted-URL explosion risk
- Server response time and its breakdown
Reading a finished report
The report opens on totals, then a fix list grouped by problem type rather than by URL. Forty broken links become one entry with one fix, not forty lines.
Every entry carries a severity, a count and a specific instruction. The instructions are pre-written for each check, so the same failure always gets the same, tested advice. The full list of affected URLs, with status codes, sits in the Excel export.
The example on the right uses made-up numbers for an imaginary shop, to show the layout. Your own report shows your real pages.
- Critical
Broken links
14 broken links across 6 pages.
Fix or remove each link, or redirect the missing page to its closest replacement.
- Critical
Console errors
3 pages log a JavaScript error on load.
Open the page with your browser's developer console and fix the script named in the error.
- Warning
Duplicate meta descriptions
11 pages share 2 descriptions.
Write one description per page that says what that page offers.
Why you can act on the findings straight away
Because nothing in the report is an opinion. Watch Owl records what happened and shows you the evidence.
Measured, not guessed
A finding is a status code, a console message, a failed axe-core rule or a DNS record. No language model reads your site and decides what's wrong, so an unchanged site gives the same findings on a re-crawl.
Safe on a live site
It never completes a purchase, never logs in, skips password fields and leaves outside sign-in pages alone. It reads your site; it doesn't change it.
Fixes you can hand over
Each fix is written for the person who will do it. Send the Excel export to a developer and they get every URL, status and error in one file.
Add your own data for a second layer
A crawl tells you what's broken. Connecting your accounts tells you which broken pages people actually visit.
| Connect | What it adds to a crawl |
|---|---|
| Google Analytics 4 | Engagement and bounce figures beside each crawled page |
| Google Search Console | Google's own URL Inspection for any crawled page, and whether your sitemap is submitted |
| Microsoft Clarity | Session behaviour metrics for your pages |
| Nothing to connect | A PageSpeed (Lighthouse) score for any crawled page, mobile or desktop, in 5 to 30 seconds each |
You connect each account once on the integrations page, and the other agents that use the same data share that connection.
Where Watch Owl stops
It's a careful inspector with clear limits. Knowing them saves you a surprise.
- Your own links only. Links to other websites aren't tested, because many sites answer automated checks with errors even when the link works. For one page's outbound links, the single-page link check tests internal and external links and tells you which sites refused to answer.
- Test messages will arrive. Ordinary forms really are submitted, from “QA Test” at [email protected], so expect those in your inbox or CRM after each crawl.
- 500 pages per crawl. Larger sites get their first 500 pages reached through internal links.
- One website per account. Re-crawl it as often as you like, up to 5 crawl starts an hour.
- It runs when you start it. There's no automatic schedule, so run it after each release or content update.
- Public pages only. Anything behind a login isn't reached.
- It reports; you fix. Watch Owl doesn't edit your site or your code.
- Results are kept for 30 days, with your 4 most recent crawls saved.
The whole crawl on one map
Five layers, from a single page up to what you do next. The linked leaves open the related tool or agent.
A whole-site crawl or a quick single-page test?
Use Watch Owl when you want the whole site checked. Use the single-page tools when you need one answer in seconds, without an account.
| Watch Owl | Single-page tools | |
|---|---|---|
| Scope | Up to 500 pages in one run | One page or one URL list |
| Checks | 54, page-level and site-wide | One job each: links, redirects, indexing, mobile, schema |
| Links to other sites | Not tested | Tested by the link check |
| Account | Needed | Not needed |
| Output | Report, fix list, Excel export, history | On-screen result |
The single-page tools: redirects, indexability, mobile-friendliness and schema. For a search-focused review of how your site is built, see Owl-gorithm Hacker.
Questions people ask before their first crawl
How often should I run a crawl?
After anything that could break a page: a release, a theme or plugin update, a redesign or a batch of new content. You can re-crawl your site as often as you need, up to 5 crawl starts an hour, and each run is compared with the last one for visual changes.
Will it send messages through my contact form?
Yes, once per crawl for each ordinary form it finds. It types test values such as "QA Test" and [email protected], so you can filter those out of your inbox. It never presses a pay, buy now, place order or checkout button, and it never submits a form on a page that looks like checkout or payment.
Does Watch Owl check links that point to other websites?
No, only links between your own pages. Social networks and many large sites block automated requests and answer with error codes even when the link works for a person, which would fill your report with false alarms. To test one page's outbound links, use the single-page link check on our tools page.
How long does a crawl take?
It depends on how many pages you have and how fast your server answers, so there's no fixed time. Results appear while the crawl runs, and if you stop it early, everything found up to that point is kept.
Can I run it on a client's or a competitor's website?
No. A CMO Owls account works on one website, which locks to the account the first time you use it. If that domain already belongs to another account, the crawl is refused.
Does it work on sites built with React, Webflow, Shopify or WordPress?
Yes. Watch Owl loads every page in a real Chromium browser and lets its JavaScript run, so it sees what a visitor sees, whatever built the site. It only crawls public pages; it doesn't log in.
Does Watch Owl use AI to decide what's wrong?
No. Every finding is something it measured, such as a status code, a console message or a failed accessibility rule, and each fix in the list is a pre-written instruction matched to the check that failed.
Related agents
Related reading
Find out what's broken today
Enter your address, create your account and the first crawl starts.