4 SEO tools paywalled the same data. The free Google tool found a bug instead.
Quick context: I test AI/SEO tools firsthand against my own real site (solostacklab.net) before writing anything about them - actual signups, actual paid trials where required, not just reading the pricing page. Disclosure since it's relevant here: some of the tools below are ones I have affiliate links to on my site elsewhere, for the parts of them that are genuinely useful. This post isn't about avoiding them, it's one specific pattern I noticed after actually using all four.
The pattern: Surfer SEO's Auto-Optimize button, SE Ranking's Site Audit page-level detail, Semrush's Core Web Vitals widget, and CoSchedule Headline Studio's SEO Score tab all do the same thing. The free tier or trial shows you a real number or tells you something's wrong, then asks for money the moment you want the part that actually fixes it. Four different vendors, same exact shape - and I only noticed because I signed up and pushed each one far enough to hit the wall myself.
Core Web Vitals specifically was paywalled in three of those four. So I checked whether Google's own free PageSpeed Insights covered the same ground - no signup, no card, official tool.
It had a category I hadn't seen before: Agentic Browsing (labeled experimental), which checks things like llms.txt compliance and whether AI agents can actually browse your site cleanly. My homepage passed 3/3. One of my review pages scored 2/3 - flagged for Cumulative Layout Shift at 0.154.
Root cause: every screenshot in my reviews was a plain markdown image with no width/height set, so the browser couldn't reserve space for it before it loaded, and the page jumped as each one popped in. Checked how widespread it was - 29 images across all 10 published reviews had the same gap, not just the one page the tool happened to flag.
Wrote a script to read each image's real pixel dimensions and fix all 29 at once instead of patching one page. Redeployed, reran PageSpeed Insights against the same URL: CLS went from 0.154 to 0, Agentic Browsing back to 3/3.
Small bug, but it's the same shape as the last post here (og:image tags I'd marked "done" that weren't) - a real gap I wouldn't have caught just looking at the pages myself, caught by a completely free tool that happened to cover ground four paid ones were gatekeeping.
Full writeup with the before/after screenshots: solostacklab.net/reviews/pagespeed-insights/
Curious if anyone else has run into this same "free tier shows the problem, paywalls the fix" shape with other SEO tools - feels less like a coincidence the more of these I actually pay for and test.
This is a good reminder that expensive SEO tools aren’t always the first place to look. PageSpeed Insights catching a real CLS issue across 29 images is a great example.
We’ve taken a similar approach with ScaleBlogger, use Google’s own data first, then bring in paid tools where they actually add value.
yeah, that ordering seems right - free/official signal first, paid tools only where they actually add something beyond it. the CWV case was a clean example since three of the four paid tools weren't even measuring the same thing PSI was.
This is a really useful point. A model saying something was blocked or handled correctly is very different from having actual execution evidence to prove it.
For AI agents using multiple tools, do you think the harder problem is catching the failure in real time, or tracing the exact step where things went wrong afterward?
catching it in real time, based on what we've actually run into. the Make failure didn't announce itself - the scenario log said "success," the email just never sent. tracing backward was actually the easy part once we noticed something was wrong (checked the inbox, worked back to the OAuth scope). the hard problem is that a lot of these tools report status optimistically, so "it says success" and "it actually worked" aren't the same claim, and you don't know to go looking until something downstream tips you off.
Skip the paywall. Ask five buyer questions in a fresh chat and log you, competitor, or nobody. That weekly count beats another tool screenshot.
that's a cheaper signal than what we've been running - SE Ranking's AI Search and Semrush's AI Visibility both reported 0% citation presence for this domain, but that's tool-reported, not something checked by hand. five buyer questions a week in a fresh chat is something we can actually track without another dashboard. will start logging it.
Disclosure first: I build an SEO tool, so I have a stake in how this shape gets described.
I do not think it is coincidence. It is where marginal cost jumps. Detection is nearly free: PageSpeed, CrUX and a crawl hand you the number without the vendor doing much per customer. Remediation needs write access to the site, a recommendation engine someone maintains, or human judgement, and all three cost real money per user. So four vendors independently land the paywall on the same boundary, because that is the first place their costs start scaling with you.
One caution on your result. PSI is lab data from a single simulated load, and CLS is particularly viewport and connection dependent. Going 0.154 to 0 in Lighthouse does not guarantee field CLS moved. Worth checking the CrUX 28 day number before calling it fixed, because image dimensions usually do fix it for real, but you want the field data to say so.
Did any of the four paywall the detection itself, or was it always the fix?
Good catch to double-check before I answer - went back through all four reviews and the "Core Web Vitals" framing in my post was too broad. Only Semrush actually paywalled Core Web Vitals specifically (the widget itself, not just the fix - "Available with a paid plan," no number shown at all). The other three paywalled different kinds of audit detail unrelated to CWV (Surfer: a content-optimization button, SE Ranking: general site-audit issue detail, Headline Studio: its SEO Score/suggestions tab). The real common pattern is "paywalled audit detail" broadly, not CWV specifically - should have been more precise. Went and fixed the review + tools data on the site to match, not just noting it here.
On your economic theory: Semrush's case actually cuts against "detection is free" a bit, since they paywalled the number itself, not just the fix. Interesting that the one CWV-specific case didn't even give away the free number.
On CrUX: honestly hadn't checked, and probably should have flagged this - the domain almost certainly doesn't have real 28-day CrUX data yet given how little traffic it's getting so far. Lighthouse-lab CLS going to 0 is the only signal I actually have right now, not field data. Good flag, will note that caveat.