I audited my own SEO properly for the first time. I'd been wrong about three things.

Your closing habit — open the results and look at who's actually there — has a direct equivalent for the LLM-citation hope in your third point, and almost nobody runs it. Ask the models the comparison question your page targets and see whether you get named. It's the same thirty seconds, and it's a check rather than a hope.

I'd temper the optimism though. I ran a daily agent publishing GEO-targeted articles for two weeks — 18 pieces, structurally correct, every claim sourced and dated. Traffic was zero and so was revenue. The execution wasn't the problem; nothing pointed at any of it. That's the part I'd flag for your comparison pages: models do lean on that format, but they still need a reason to weight your version over the fifty other pages with the same title. Zero impressions and zero inbound links are signals to a retrieval system too, not just to Google.

So leaving the pages up is fine — they cost nothing to keep standing, and you're right that it's the format that gets pulled into comparison answers. I just wouldn't count it as the reason it was worth writing until you've asked the question and seen your own name come back.

The deeper issue here is that the SEO mistakes all come from optimizing the page in isolation rather than the searcher’s decision context. “Pinterest fonts” may be the better keyword, but ranking for it only matters if the resulting visitor sees a page that immediately matches the reason they searched that phrase. That’s the conversion layer underneath the SEO audit: search intent gets you into the SERP, but message-to-intent alignment determines whether the click becomes meaningful. I’d be curious whether the retitled pages are being evaluated not just by rankings and traffic, but by what visitors actually do once they arrive.

The "referring domains lying in both directions" point is the one most people never catch. I've seen founders give up on a keyword because the difficulty score scared them off, when the actual SERP was three thin affiliate pages that would fall to a decent article. And the reverse: chasing a "low difficulty" term that's really just riding domain authority nobody else has. Manually eyeballing the SERP before committing hours to a page is such an unglamorous habit, but it's the only step in your list that can't be gamed by a metric. Out of curiosity, are you tracking this stuff (rankings, backlinks, which pages to retitle) in a spreadsheet, or does it live in your head until you do an audit like this one?

The comparison content point deserves one nuance, because I think the rule is narrower than "comparison articles need brand awareness." What fails pre-traction is comparing yourself to a competitor, since nobody searches for a brand they've never heard of, and yours is the one they've never heard of. Comparing two famous things that aren't you works from day one, because you're borrowing their search demand instead of relying on your own. My first blog post is a comparison of two AI tools people already argue about (I teach non-coders to use one of them, at kambolab), and that choice was made precisely because my own name has zero volume. Same format, opposite mechanics.

Your closing line is the keeper though. "Trusting a number that was easier to get than the right one" describes most of my early decisions, well beyond SEO.

This is a good reminder that SEO metrics can create a false sense of certainty. ~

Search volume and difficulty are useful signals, but seeing who actually ranks tells you much more.

The keyword example is a great one. Four times the volume with a fraction of the competition sounds like an obvious choice until you look at the actual SERP.

I also liked the point about comparison content. It’s easy to write what seems like a logical SEO topic without asking whether people have any reason to search for it yet.

The “trust the number because it’s easier” pattern feels like the biggest lesson here.

Checking the actual results first seems like a small habit that can prevent a lot of wasted work.

those are two different queries though. "myproduct vs competitor" does need brand awareness, but "[competitor] alternatives" doesn't, the searcher not knowing you is the whole point of that query. we run affiliate software on stripe and worked that serp twice, eleven days apart, and almost every result was either a competitor's own alternatives listicle or a review/directory site, so the lever is getting listed on the pages that already rank rather than publishing your own. even that has a wall, the one genuinely independent listicle we found says it won't take sponsorship or placement requests. so the format isn't dead, your page just wasn't the kind that ranks for it.

The pivot from high-volume, high-difficulty keywords to those that actually align with your site's current authority is the most undervalued step in SEO. We often get caught up in vanity metrics or 'obvious' search terms, but the reality is that chasing a keyword with a 26-domain-backlink barrier when you are starting out is just burning cycles. Focusing on lower-competition, long-tail terms might feel like you're aiming smaller, but it is the only way to build the steady traffic floor you need to eventually rank for those head terms.

The comparison content pre-traction mistake is one I watch founders repeat. I did it too. That keyword requires people to already know you exist. The SERP tells you in 30 seconds: if every result is a third-party review site, you're not getting in without the brand recognition to match. The insight on difficulty vs referring domains is the important one. A difficulty 14 keyword I chased turned out to be a SERP full of Shopify subdomain pages. Difficulty scores break in both directions depending on who's actually ranking. Have you found a tool that does the SERP check automatically, or is it still manual?

Finding three assumptions were wrong once you looked at the underlying evidence is the interesting part.

Has the audit changed just what you’re doing in SEO, or has it changed how you decide which growth assumptions are trustworthy enough to act on in the first place?

The "vs competitor" point hit hardest — did the same pre-traction, got the same zero. The part that clicked: comparison pages don't just need volume, they need people to already know the frame of reference. "MyProduct vs X" requires brand awareness; "tool that does Y" doesn't. One is post-traction SEO, the other is discovery SEO. We now build every page around the discovery phrasing and treat comparison content as a later asset, not a launch play.