Most owners who call us say some version of the same thing: "we already tried SEO, it didn't work, why would this be different." It's usually said with a mix of frustration and embarrassment — like they got fooled once and don't want it to happen twice. I don't argue with that feeling. It's earned. What I do instead is ask three questions, in order, before I say a single word about what we'd do differently. In almost every case, the answer to one of those three questions explains the whole story — and it isn't "SEO doesn't work."
That's the conversation this article is. Not a pitch for why our version is better. An honest walk through the three failure modes that account for nearly everyone who's ever told me SEO didn't work for their business, plus how to tell which one happened to you before you write off the entire discipline — or before you hire someone new and repeat the same mistake with better branding.
"SEO didn't work" almost always means one of three things happened: the work was a single project instead of an ongoing habit, keywords were researched but no real page was ever built and shipped for them, or a redesign quietly broke technical fundamentals nobody checked before-and-after. None of those three is evidence that search optimization itself is broken — they're evidence of how it was run.
Failure mode one: the one-time project
This is the most common story by far. An agency or freelancer was hired for three to six months, did a round of on-page fixes, wrote some content, built a few backlinks, and then the engagement ended — because that's how the contract was scoped, not because the work was finished. Nobody logged back in six months later to see what happened next.
Here's the problem with that structure: Google's own SEO Starter Guide frames search optimization as an ongoing practice tied to how your site and your competitors' sites keep changing — not a task you complete once. Rankings aren't a bank account you deposit into and then withdraw from forever. They reflect the current state of your site against the current state of everyone else's, re-evaluated continuously. A site that stops improving doesn't hold its position; it slowly loses ground to competitors who kept publishing, kept building links, and kept fixing technical issues after the invoice was paid.
The tell that this is what happened to you: pull up your analytics and look at the traffic trend line during and immediately after the engagement. If it rose while the work was active and then flattened or declined the month the contract ended, that's not proof SEO failed — that's proof the growth was rented, not owned, and the lease ran out.
Failure mode two: keywords with no page behind them
The second most common story: someone handed you a spreadsheet of keywords — sometimes a genuinely good one — and nothing happened after that. No page got built. Or a page got built, but it was 300 words stuffed around the keyword instead of an actual answer to the question a real person was asking.
Keyword research identifies demand. It does not create supply. If the deliverable you received was a list of terms and search volumes, you were handed half the job. The other half — building a page thorough enough, clear enough, and specific enough that Google's ranking systems and, increasingly, AI engines actually want to serve it — is where almost all of the real work and most of the real cost lives. A list of keywords with no page is a map to a city with no city built there yet.
Check this by opening the keyword list you were given (if you still have it) and searching your own site for each term. If most of those keywords have no dedicated page — or the page that exists doesn't actually answer the specific question in the keyword — that's your failure mode. The research wasn't wrong. It was never finished.
Failure mode three: the redesign that broke it quietly
This is the one that's hardest to trace back, because it looks like nothing happened, when in fact something specific and fixable broke. A new website launched — new design, new developer, new CMS — and organic traffic slid over the following months. Nobody connects the two events because the decline showed up gradually, not the day of launch.
What actually happens in these redesigns, almost every time: URLs change without 301 redirects, so every backlink and every bit of accumulated ranking signal now points at a dead page. Internal links get rebuilt around the new navigation and stop pointing at the pages that used to rank. Page speed regresses because the new theme ships more JavaScript. Sometimes the new site renders its text client-side in a way that search and AI crawlers can't read at all — a problem we cover in full in why AI crawlers can't see your website.
The tell here is timing, not content. If you pull up Google Search Console's Performance report and filter by date, a drop that starts within a few weeks of a relaunch — even if nobody "did SEO wrong" — is very likely a technical migration issue, not evidence the underlying strategy was flawed. This is exactly the scenario in why your best-performing page is invisible to AI crawlers: pages that used to work quietly stop being reachable, and nothing in the interface tells you that happened.
The honest exception: sometimes the work really was bad
I'm not going to pretend every failed SEO engagement fits neatly into those three buckets and the practitioner did nothing wrong. Sometimes the work itself was low quality — thin, templated content with no real expertise behind it, or link-building tactics that risk the kind of penalty Google's spam policies explicitly warn against. If you were sold "we'll build you 500 links a month," that's a red flag worth naming directly, not a nuance to soften. The difference between a legitimate strategy poorly maintained and a bad strategy executed well is worth being honest about, because the fix is completely different in each case.
Why the timeline matters more than owners expect
Even when the strategy and execution were both sound, a lot of "it didn't work" stories are actually "it didn't work fast enough, so we quit before it could." Google's own Search Central team has said publicly, repeatedly, that meaningful ranking movement from real changes can take four months to a year to fully show up — and that's assuming the change stays live the whole time, which circles back to failure mode one. A 90-day engagement judged on 90-day results is being graded on a curve the discipline itself was never built to pass.
The part that's actually new: AI changed the scoreboard, not the game
There's a real reason this conversation is more common in 2026 than it was a few years ago, and it isn't that classic SEO stopped working — it's that a second scoreboard appeared alongside it. Google's own AI features documentation confirms that AI Overviews are generated from the same organic search index your classic SEO work already targets — so the technical and content fundamentals still matter, arguably more than before, because they gate whether an AI system can read and trust your page at all.
What's genuinely different is that ranking #4 on a results page used to still get you clicks. Being one of three names an AI assistant mentions, or being left out entirely, is a much harder cutoff. We've written about that shift directly in the traffic didn't drop, the clicks did — a lot of "SEO stopped working" stories are actually "SEO kept working exactly as designed, and a different surface started answering the question before the click ever happened." That's not a reason to abandon the fundamentals. It's a reason to add a second discipline on top of them, which is what generative engine optimization actually is — not a replacement for SEO, an extension of it.
What to actually check before you hire anyone again
If you're standing at the point of deciding whether to try again, run this before signing anything:
- Pull the traffic trend line from the last engagement, month by month, through today. Identify exactly when it changed and what else changed that same month — a contract ending, a redesign, a CMS migration.
- Audit the deliverables you actually received. Was there a finished, published page for every keyword on the list, or just the list? Were technical issues actually fixed, or just identified in a report?
- Check whether the site is still crawlable and fast today. Run it through the Rich Results Test and view the raw HTML source — if your visible content isn't in the raw HTML, no ranking strategy can save you until that's fixed. See why AI crawlers can't see your website for the exact steps.
- Ask any new agency for the maintenance plan, not just the launch plan. If the proposal is a fixed-length project with no ongoing component, you're about to repeat failure mode one with a different logo on the invoice.
The honest version of "does it work"
SEO, run as a genuine ongoing practice, still works — Google's index and its AI features are both still built on crawling, understanding, and ranking real content, and that hasn't changed. What breaks it, almost every time we're asked to look at a past failure, is one of the three patterns above: the project ended before the compounding started, the research was never finished into real pages, or a later change quietly undid the technical foundation without anyone checking. Name which one happened to you, fix that specific thing, and judge the result on the actual timeline the work needs — not the one the last invoice was scoped to.
Questions people ask
Almost always one of three things: the work was a one-time project instead of an ongoing habit, keywords were researched but no page was actually built and shipped for them, or a site redesign broke technical fundamentals (crawlability, page speed, internal links) without anyone checking traffic before and after. Google's own Search Essentials documentation is explicit that ranking is earned continuously, not banked once and kept.
Google's Search Central team has publicly said meaningful changes can take four months to a year to show up, and that's for changes made correctly and left in place. If an owner ran a 90-day engagement, saw nothing, and quit, that timeline was very likely too short to judge the work, not proof the work was wrong.
No, but the scoreboard changed. Classic SEO still governs whether Google can find and rank your pages at all, and Google's own AI Overviews documentation confirms it draws from the same organic index. What's new is a second, parallel game — whether ChatGPT, Perplexity, and Gemini cite you directly — that classic keyword-and-backlink SEO alone doesn't win.
Want us to look at yours?
Want to know if AI is recommending you or your competitor? Get an AI visibility audit from Acromatico.
Get a free AI Visibility Audit →