A regex written with a literal backspace character instead of a word boundary escape will match nothing, silently, every time. Here is how we found four audit checks that were always returning clean results — and what it means for trusting any automated SEO tool. Content:

The problem with silent failures

Most software bugs are obvious. The page breaks. The form throws an error. Something visibly wrong happens and you investigate. The hardest bugs to catch are the ones that produce a result — just the wrong one — with no indication that anything is amiss.

A check that always returns zero issues is indistinguishable from a page that genuinely has zero issues, unless you test it against a page you know has problems.

What we found

During a recent deep audit of our own platform code we found four checks in the image and input validation tools that were always returning zero results regardless of what was on the page being audited.

The cause was a single character. A word boundary in a regular expression — normally written as \b — had been stored in the source file as a literal backspace byte (0x08) rather than the two-character escape sequence. To a text editor it was invisible. The file loaded without errors. The JavaScript parsed without errors. The function ran without errors. It just never matched anything.

The affected checks were:

Every audit that ran these checks was returning a clean pass on all four, regardless of how many images were missing alt text or how many inputs were missing IDs. Users auditing pages with genuine accessibility and performance issues were being told there were none.

How we caught it

The check was exposed during a deep code audit that tested each regex against known-bad input rather than simply parsing the source and checking for syntax errors. We wrote a test that created a sample page with images deliberately missing alt attributes and ran the checker against it. When it returned zero issues on a page we knew had issues, we investigated.

Tracing back through the source, Python's byte-level file reader showed 0x08 at the exact position where \b should appear. The fix was straightforward: replace all literal 0x08 bytes in the file with the correct \b escape. All four checks immediately began working correctly.

What this means for automated SEO tools

Automated SEO tools are only as good as their test coverage. A tool that has never been run against pages it is supposed to fail on has no evidence that its passing results mean anything.

When evaluating any SEO tool — including ours — the right question is not just "does it flag issues when I have them" but "have I verified it flags issues on a page I know has them". Create a test page with known problems: an image with no alt text, a title tag that is too long, missing meta description, no H1. Run it through the tool. If it passes clean, the tool has a problem.

We run this kind of regression testing after every significant code change. The March 2026 update includes fixes for these four checks, and we have added the test cases to our audit process so the same class of silent failure cannot recur undetected.

If you ran image or input audits through our platform prior to this update and received clean results, we recommend re-running them. The results may now be different.