Skip to content

Smart Audit

The Smart Audit screen, the difference between site-wide and page-level issues, how to read severity, and how to use a generated fix.

Site healthUpdated 26th August 20265 min read
  • the crawl, its issues, and the fixes
  • one piece of content, checked before it ships

Overview#

Smart Audit crawls your public pages and asks whether an AI engine could read them, work out what they are about, and quote them with any confidence.

There is real overlap with a traditional SEO audit, but the question underneath is a different one. SEO asks whether a page will rank. Smart Audit asks whether an engine wanting to quote that page would actually be able to.

Nothing gets installed for this. The audit runs against your public site from the outside, the same way an engine would.

Panels and controls#

Score Trend shows which way things are going, and as with everything else in Seerly the line is worth more than any single reading of it. Issues by Severity gives you a working order, since severity here reflects how much a given problem interferes with an engine's ability to use the page, so working top down is genuinely the right approach.

The screen then splits into two halves, and understanding the split is the most useful thing you can take from this page. Site-Wide Issues are problems that apply right across your domain, where a single fix resolves them everywhere and the return on your time is enormous. Page Breakdown lists individual pages and what each one failed on. Clear the site-wide list first, every time. Fixing the same underlying problem on forty pages one at a time is work nobody needed to do.

Checked URLs and All Pages tell you what the audit actually managed to cover, which is worth a glance because how much of your site gets crawled depends on your plan. Failing Pages narrows things to pages with problems, and Grouped collapses the list by issue type so you can see how widespread something is instead of scrolling past the same finding forty times. All sections and Clear filter control what is on screen, and if you hit No pages match your search it is almost always a filter still applied rather than a suddenly spotless site. A page with nothing wrong shows All checks passing.

The audit loop

your moveSeerly runs this
  1. 1

    The crawl scores your site

    Issues land ranked by severity

  2. 2

    You clear the site-wide list

    One fix resolves it everywhere

  3. 3

    You work the page issues

    By severity, key pages first

  4. 4

    You re-run

    Confirming the fix landed rather than assuming

the score trend records whether the loop is working

The re-run is the step people skip, and it is the only one that turns a fix from a belief into a fact.

Reading an issue#

Opening an issue gives you Issue Details, which covers what was found under Result and Details, exactly where it sits under Found on page and Element Location, and why it matters to an AI engine specifically under Rationale.

Element Location is the field that saves the most time. Most audit tools will tell you a page has a problem and leave you to hunt for it; this tells you which part of the page to open.

Issue Details

Missing structured data

Result
Failed
Details
No Product or FAQ schema found on the page
Found on page
/pricing
Element Location
<head>, after the canonical tag
Rationale
Without schema an engine has to infer what the page offers from prose, and often gets it wrong
Illustrative. Element Location is what turns an audit finding into a fix you can hand to a developer.

Generated fixes#

Generate Recommendation produces a suggested fix for an issue, labelled AI Generated. You see Generating... while it works, and the result arrives as Recommended Fix with a note to Review the details and apply the suggested fix.

Read the rationale before you apply any of it. A generated fix is a strong starting point, but it knows nothing about your template, your CMS, or why a developer three years ago did something that looks strange but was load-bearing. Treat it as a suggestion to a person rather than a patch to apply. Issues move through Pending and Failed as you work, so the list stays honest about what is still outstanding.

The llms.txt file#

llms.txt is a file at the root of your site describing what your site is and where the important content lives, in the way robots.txt describes crawl permissions. It is an emerging convention rather than a settled standard, and not every AI system reads it today. The honest case for publishing one is that the cost is small and the downside is nil.

You reach it from this screen: open the llms.txt issue and follow the link in its details panel. If no llms.txt issue is outstanding, that is itself the answer.

Generate llms.txt produces a draft from your own site, so you are not writing one from a blank page. You see In Progress while it works, then Complete, or Generation Failed. Regenerate runs it again after your site changes. Select a file to preview shows an individual page's entry.

Read the descriptions before publishing. These are the sentences you are handing to any system that reads the file, so they should be what you would want quoted back at you. Aim for an opening line that states plainly what you do: "We make X for Y" can be quoted, "reimagining the future of Z" cannot. Link to pages that genuinely answer questions rather than campaign landing pages, and keep it current, since a file describing a product you have since changed is worse than no file.

The draft is also a finding in its own right. It is built from what Seerly found on your site, so it approximates what an outside reader concludes about you from your pages. If it describes your business oddly, that is not a generation problem. It is roughly what an AI engine reading your site would conclude. A page showing No content available means the crawler could not extract anything from it, which usually points at a rendering problem in the site-wide list above.

Publishing the file does not make an engine cite you, and adoption is uneven. Treat it as one low-cost signal among several. It is also no substitute for the fundamentals: if crawlers cannot reach your pages, a file pointing at them changes nothing.

Best practices#

Crawler access comes first, and it is worth treating as a prerequisite rather than the first item on a list. If AI crawlers cannot reach your content, nothing else on this screen can help you. Verify it explicitly instead of assuming, because a site can quite happily serve a complete page to Google and a near-empty shell to everything else, which sails through conventional SEO checks while remaining invisible to AI engines. Look for crawler access and rendering problems in the site-wide list and treat anything you find there as blocking.

Once that is settled, what you are aiming for is pages that answer one clear question with specific, quotable claims. A page that hedges is hard to cite no matter how good its markup is.

Next steps#

  1. Let the run finish. Mid-run counts are incomplete rather than encouraging, and a partially populated audit can look remarkably like a clean bill of health.
  2. Fix crawler access before you touch anything else.
  3. Clear Site-Wide Issues next, where one change covers every page you have.
  4. Use Grouped on the page breakdown to find the issue hitting the most pages at once.
  5. Work by severity, and within a severity band, start with the pages you most want cited for your key topics.
  6. Re-run after making changes to confirm the fix landed rather than assuming it did.

Limitations#

Smart Audit checks whether your content can be read and cited. It has nothing to say about whether that content is persuasive, or whether it answers the questions your buyers are actually asking. Citations is where you find that out.

How much of your site gets crawled depends on your plan's page cap. See seerly.app/pricing.

Passing every check means engines are able to use your content. It does not mean they will.

Related