I run SEO for a small portfolio of three content sites, and every quarter I review the top 20 articles on each for on-page improvements. For years my 'process' was: open the article, open the SERP top three, compare them in my head, and write down whatever jumped out. The problem with reviewing on vibes is that vibes are inconsistent. Some weeks I'd obsess over E-E-A-T signals and completely forget to check internal links. Other weeks I'd fixate on structure and never notice the FAQ section was missing schema. Same auditor, wildly different audits, depending on what I'd had for lunch.
The thing about inconsistent auditing is that it feels productive while you're doing it. You open a page, you spot a problem, you fix it, you feel like an SEO. But you're playing whack-a-mole with whichever mole happens to be sticking up that day, and the moles you fixed on page one are quietly popping back up on pages three through fifteen because you forgot to check them there. I had no idea how bad my coverage drift was until I forced myself onto a fixed checklist and saw the same four issues staring back at me from every single audit.
The same 20 factors, every time, no exceptions
The On-Page SEO Audit — Score Any Page Against 20 Ranking Factors prompt gave me the one thing my brain refused to provide: consistency. I run each article through it on Grok 4.3 — fast, punchy, and it doesn't pad the output with throat-clearing, which is exactly what I want when I'm grinding through 15 pages. The output is a clean table of PASS / FAIL / REVIEW with a specific 30-minute fix bolted onto each factor. No essays. Just the diagnosis and the prescription.
Here is the part that made me sit up. Across 15 articles on one site, the same four problems appeared in literally every audit. Every one. No FAQ section despite the queries being question-shaped. Internal links using lazy anchor text like 'click here' or 'read more' instead of the target page's keyword. The keyword showing up dutifully in paragraph one and then vanishing for the next 400 words. And threadbare E-E-A-T — no author bio, no publish date, no cited sources. I had been missing all four, intermittently, for a year, because my manual review never checked the same things twice.
The systemic fix, in one sprint
- Added schema-eligible FAQ sections to 12 articles in a single dev sprint
- Rewrote internal-link anchor text site-wide to use each target page's actual keyword
- Reintroduced the keyword into mid-article H2s on 10 articles so it didn't disappear
- Added author bios and publish dates to all 15 via a template update
Two months later the numbers had moved: average position across those 15 articles improved from 8.2 to 5.9, with CTR rising alongside it. The new FAQ schema earned featured snippets on four of the twelve articles I'd added it to. None of this was a clever new tactic — it was four boring, well-known fundamentals that I'd been applying inconsistently because I'm a human and humans drift.
Where the prompt is honest about its blind spots
I want to flag the thing I respect most: the prompt does not bluff. It cannot measure raw page speed or validate structured data from pasted text, and instead of inventing a confident-sounding lie, it stamps those factors [NEEDS A CRAWL TOOL] and moves on. That honesty is what lets me trust the factors it does evaluate. A tool that pretends to know everything is a tool you have to second-guess on everything.
There's a second-order benefit I didn't expect: the audit output is also a teaching tool. When I hand a writer an audit that says their internal-link anchors are weak and shows them the exact fix, they internalize it, and the next article they submit already has better anchors. After two quarters of running every article through the same 20-factor lens, my writers started pre-empting half the findings on their own, because the checklist had effectively trained them. The audit stopped catching the same four problems not because I changed the prompt, but because the team stopped making them.
The real value here isn't novelty, it's discipline — the same checklist, applied identically, every single time, catching the gaps my attention span lets slip. I pair it with my SERP Snippet Lab — Three Meta Title + Description Variants Tuned for CTR, Trust, and Keywords prompt so a weak title and a thin page get fixed in the same review pass. Grab the On-Page SEO Audit — Score Any Page Against 20 Ranking Factors prompt on Prompt Dock and run it across your ten best articles — I promise you'll find the same problem in more than one of them, and finding the pattern is the whole point.