What audit tools can and cannot see

I run audits on my own sites every week, and I’ve stopped treating the output as a verdict. It’s a set of measurements taken by a machine that visits a page under specific conditions, and those conditions don’t match what a real visitor or Google’s own systems experience. That gap matters more than most people admit, especially when a client asks why the tool gave them a 92 and their traffic still dropped.

This isn’t an argument against using audit tools. I use three or four of them regularly and I’d be flying blind without them. It’s an argument for knowing exactly what each number is built from, so you stop treating a crawl report like ground truth.

A crawler is a guest, not the owner

Every audit tool works the same basic way: it sends a bot to fetch a URL, reads the response, and follows links it finds in the HTML. That’s it. The bot respects (or ignores, depending on settings) your robots.txt file, requests pages the way a browser would, and logs whatever comes back: status codes, header tags, word counts, load times, internal and external links.

What it does not get is anything that depends on being Google. It doesn’t see your Search Console data, your actual crawl budget, your site’s history, or how Google’s indexing systems have treated your domain over time. A third-party crawler is a separate visitor with its own limited view, not a window into Google’s index.

Where the technical checks are genuinely solid

Broken links, missing alt text, duplicate title tags, redirect chains, orphan pages, thin content by word count, malformed structured data. These are things a crawler can catch reliably because they’re objective and visible in the raw response. If a page returns a 404, every tool will agree it’s a 404. If two pages share an identical title tag, that’s a fact, not an inference.

This is the layer where audit tools earn their keep. I trust this output almost completely, because it’s not modeling anything, it’s just reading what’s there.

Rendering is where a lot of tools quietly fail

Most sites now load at least some content with JavaScript. A crawler that only reads the initial HTML response will miss anything injected after the page loads, unless it runs a headless browser to render the page first, the way a real user’s browser does. Some audit tools do this by default, some do it only on a slower “JS crawl” setting, and some skip it to keep crawl speed up.

The practical effect: a tool can flag a page as “missing H1” or “thin content” when the content is actually there, just rendered client-side after the crawler already moved on. I’ve had this happen on a site rebuild where the audit screamed about missing body copy that was fully visible in a browser. The fix wasn’t the content, it was re-running the crawl with rendering turned on. If you don’t know whether your tool renders JS by default, you’re trusting a number without knowing what it’s measuring.

Backlink data is a sample, not a census

This is the one people misunderstand most. Ahrefs, Semrush, Moz, and similar tools each run their own independent web crawlers to build their link indexes. None of them have access to Google’s link graph. Their “referring domains” count is a function of how much of the web their bot has crawled, how recently, and how deep it went on any given site.

Two tools will show different backlink counts for the same page because they crawled different slices of the internet at different times. Neither number is “the real one.” Both are partial views built from separate crawls. When a tool reports zero backlinks to a page, that can mean the page truly has none, or it can mean the tool’s crawler hasn’t found them yet.

This also means a tool cannot tell you whether a link came from a paid placement, a private network, or a genuine editorial mention. It can show you the anchor text, the source domain’s crawled metrics, and a spam score built from patterns the vendor trains on. It cannot see intent or ownership. A clean-looking spam score is not the same as a safe link, and I wouldn’t treat link scheme participation as low-risk just because a tool’s dashboard looks quiet. The tool has no way to know what it can’t see.

Rank tracking is a snapshot, not your reality

A rank tracker queries a search engine from a specific location, usually a data center, often logged out, with a default set of settings. It records where a URL lands for that query at that moment. Real searchers get personalized results shaped by their location, search history, device, and whatever test bucket Google happens to have them in that day. The position a tool reports and the position an actual person in a specific city sees can differ, sometimes by a lot, and neither is wrong, they’re just different queries run under different conditions.

This is also why day-to-day rank fluctuation in a tracker isn’t automatically a signal of anything. Some of it is noise from how the tool samples results. I look at trend over weeks, not single-day swings, because a single query snapshot from one data center isn’t a reliable read on how the page is actually performing across real searchers.

Lab speed scores and real user speed are different measurements

Most audit tools generate a performance score using something like Lighthouse, run once, in a controlled environment, on a single simulated connection and device profile. That’s lab data. It’s useful for diagnosing what’s slow and why, because it gives you a waterfall you can act on.

But the page experience signals that actually factor into how Google evaluates a site come from field data, aggregated from real Chrome users over a rolling window, through the Chrome UX Report. A page can score well in a lab test and still perform poorly for real users on slow connections or older devices, or the reverse. If your audit tool only shows you a lab score, you’re seeing a diagnostic, not the metric that reflects actual visitor experience.

What no tool can judge for you

Word count, keyword density, readability grade level, heading structure. All measurable, all reported confidently, none of it tells you whether the content is actually useful to the person reading it. A page can hit every on-page checklist item and still be shallow, repetitive, or wrong. A tool can count how many times a keyword appears; it can’t tell you if the page actually answers the question someone typed in.

The same is true for search intent. A tool can show you what a page currently ranks for. It can’t tell you why a searcher clicked away in four seconds, or why a competing page with fewer backlinks and a worse speed score is outranking you anyway. Those answers usually come from reading the actual search results yourself, looking at what’s winning, and being honest about whether your page genuinely competes with it.

How I actually use the reports

I run technical crawls to catch objective errors, broken links, duplicate tags, indexation problems. I use backlink tools to spot patterns and check my own link building against what competitors are doing, not as a definitive count. I use rank trackers for trend direction over time, not daily panic. I run speed audits to find what to fix, then check field data separately to see if it moved the needle for real visitors. None of these tools tells me whether the site will rank better next month, and I don’t trust anyone who claims a tool can promise that.

The audit is a starting point. It tells you what’s measurable. What’s not measurable, whether the content is genuinely good, whether a link is safe, whether a real person found what they needed, is still a judgment call, and it’s still yours to make.

If you want a closer look at how we approach technical audits, link data, and content decisions without pretending a dashboard has all the answers, head back to the SEO Desk homepage.