Why So Much SEO Work Doesn't Drive Growth Anymore (and What We Did About It)

Search Engine Land published a piece in June 2026 arguing that most SEO teams are busier than ever and still not growing — because keyword lists, high-volume content, and on-page tweaks used to be most of the job, and now they're just the starting point. We think that's exactly right, and we didn't just nod along. Over the last two weeks we ran our own site through the same test the article implies every SEO program should pass: not "is the content good," but "can an AI crawler that never executes JavaScript actually see any of it." The honest answer, when we checked, was no — for most of the site. This is what we found, what we fixed, and what we'd tell any team wondering why their SEO retainer keeps producing more content without producing more growth.

What Actually Changed

For years, ranking well meant hitting the fundamentals: relevant keywords, enough content depth, clean on-page structure. Those things still matter — they're just not where growth comes from anymore. AI Overviews, ChatGPT, and Perplexity now sit between a lot of searches and a click-through, and what gets cited there depends less on keyword density and more on whether the content is structurally extractable, backed by real authorship, and actually reachable by a crawler that can't run JavaScript. The Search Engine Land piece put it bluntly: the job those old SEO skills add up to is no longer the job that moves the needle. Most teams still have the old foundation covered. Almost none of them have built anything on top of it.

We Read It, Then Spent Two Weeks Proving It

Rather than write another "SEO is changing" post, we decided to test the argument against our own site — the same one running our own paid media business. What we found was uncomfortable enough to be useful.

• Only 6 of roughly 37 live pages were actually rendered server-side. Everything else — the homepage, About, the full services index, every blog post, every case study — shipped as an empty page shell that GPTBot, ClaudeBot, and PerplexityBot literally cannot read, since none of them execute JavaScript.
• FAQ schema was still attached to commercial pages even though current guidance treats structured FAQ markup as de-prioritized there — the visible Q&A text extracts fine on its own, the schema was just dead weight we hadn't cleaned up.
• Blog posts carried a generic company byline instead of a named author, which meant zero real E-E-A-T signal on the one content type where it matters most.
• There was no llms.txt file at all, and no visible "last updated" date on most pages — both small, both free, both skipped.

None of this needed new content. It needed the existing content to actually be visible and attributable. We rebuilt the build pipeline so every route renders full HTML at build time instead of relying on the browser, reassigned blog authorship to the actual person who wrote each piece with a real bio and LinkedIn link, dropped FAQ schema from commercial pages while keeping the FAQ content itself, and shipped an llms.txt describing every major page. Two weeks of engineering work, zero new blog posts, and the site went from mostly invisible to AI crawlers to fully readable.

The Old Checklist vs. What Actually Moves AI Visibility Now

Where to Start This Week

• Check whether your actual pages render without JavaScript — view the raw HTML a crawler sees, not what loads in a browser. This was our single biggest finding and probably will be yours too.
• Audit where your team's hours actually go. If it's still mostly keyword lists and content volume, that's the old job, not the one that moves the needle now.
• Pick one flagship piece — a real data point, a real case study, something only you can publish — instead of another generic explainer.
• Assign someone to own AI visibility specifically: crawler access, structured data, llms.txt, authorship. It's nobody's job by default, which is exactly why it doesn't get done.
• Reset what "SEO work" is being measured against. Rankings and traffic still matter, but citation in AI answers is a separate, now-real outcome worth tracking on its own.

Frequently Asked Questions