Your Technical Audit tells you how readable a page is for an AI model once that model has fetched it. It does not test whether a particular AI model can reach your page in the first place, so treat it as a readability assessment rather than a crawl check.
This article covers which pages to audit, what the results mean, and how to act on them.
Which pages should I audit?
Start with the pages AI models are already using. Improving a page a model has cited tends to pay off faster than improving one it has never touched.
Gumshoe does this shortlisting for you. Below your audited pages, at the bottom of the Technical Audit page, it suggests pages from your brand's domain that were cited in your latest Visibility Audit or picked up recommendations in your latest Content Audit, with the citation and recommendation counts shown next to each one and the most cited pages first. A page with a high citation count and open content recommendations is your strongest candidate. You can add any of them to your audit directly from that list.
Beyond that, cover your page types rather than every individual URL. A product page, an article, an FAQ page, and a landing page usually share a template each, so a fix on one often carries across the whole set.
How do I tell whether a page is worth the effort?
Click the page URL near the top of any Analysis results page to open URL details. It gathers everything Gumshoe knows about that one URL: its full technical audit history, how many times AI models cited that exact URL in your latest Visibility Audit run, which prompts produced those citations, and any Content Audit recommendations pointing at the page. The link icon next to the URL opens the live page in a new tab.
Use it to sort effort from busywork. A page with citations behind it and open content recommendations is worth fixing first, because a model is already reaching for it. A page with no citations can still be worth fixing when it uses a template you rely on across the site, but it is rarely the most urgent thing on the list.
It is also the fastest way to see whether a page has been failing for a while rather than just on the latest run, since every run is listed in one place.
What does the Technical Audit check?
The Technical Audit runs 20 pass or fail checks against a page, grouped into four categories.
- Machine readability and data layer: The page loads, is not blocked from AI crawlers, and has valid structured data, a canonical tag, a title, Open Graph and Twitter card tags, a publish date, and a mobile viewport. 10 checks.
- Content clarity and answer readiness: One clear main content region with enough text, logical headings, readable paragraph length, a healthy text to markup ratio, and emphasized key points. 6 checks.
- Authority and trust signals: An author in article structured data, and a visible author byline. 2 checks.
- Multimodal and agent optimization: Image alt text, and captions on video or audio. 2 checks.
Machine readability and data layer carries the most weight, so it is usually the biggest lever on your score.
How do I read a result?
Every check returns one of three results, and each one expands to show what it looks for, why it matters, and how to fix it.
- Pass: the page meets the check.
- Fail: the page does not meet the check. These are your action items.
- Not applicable: the check does not apply to this page, so it drops out of the score rather than counting against you. A page with no images is not scored on alt text.
Some checks only run if an earlier one passes. Structured data validity is only checked when structured data exists on the page, for example. Because pages can end up with a different number of applicable checks, your score is most reliable for tracking the same page over time rather than comparing one page against another.
What should I fix first?
Work from the failed checks, heaviest category first.
- Fix first: Failed checks that affect whether the page can be read at all, such as HTTP status, robots blocking, and JSON-LD presence and validity.
- Fix next: Other failed checks, starting with the heaviest category, such as canonical tag, title quality, heading hierarchy, and main content volume.
- Then: Failed checks in the lighter categories, such as author byline, image alt text, and media captions.
What does a good routine look like?
- Review your latest run.
- Check that the page loaded correctly before acting on any content findings. If the HTTP status check failed, or the scan details show a redirect, fix or flag that first. Content checks on a page that did not load fully are not reliable.
- Work through the failed checks, starting with machine readability and data layer.
- Rerun the page after a fix ships, to confirm the check now passes.
- Check the suggested pages list periodically and add anything new that is picking up citations, and open URL details on any page you are unsure about before spending time on it.
How should I report scores to my team or my client?
Do:
- Track the same pages over time rather than comparing different pages to each other.
- Report the count of passed and failed checks alongside the score. It is easier to explain than a single number.
- Note when a score change followed a specific fix, so the improvement is attributable.
Don't:
- Compare two different pages' scores directly without noting that they may have been scored on a different number of checks.
- Mix results from workspaces with different configurations without saying so.
Will my past scores change if a check is updated?
No. Every check carries its own version, and each run records exactly which versions and weights were used. If a check is refined later, that becomes a new version going forward. Your past results stay as they were.
Can I adjust the weights or turn off checks for certain page types?
Not currently. Every page is scored against the same configuration.
Where do I get help with a specific result?
Send us the page URL and the run in question, either by asking for a human in chat or by emailing support@gumshoe.ai, and we will review it with you.
Pro tip: Most machine readability failures are template-level rather than page-level. Fixing a missing canonical tag or an absent Open Graph block in your template can move dozens of pages at once, which is why that category is worth clearing before anything else.