AI in SEO: what actually works, and what's just noise

TL;DR

• AI in SEO is a faster way to do the slow parts (research, first drafts, briefs). It is not a new set of ranking rules.
• About 86% of SEO professionals now use AI in their work, so using it isn’t an edge any more. Using it well, and editing it hard, is.
• AI Overviews are cutting clicks. Pew found people click a normal result 8% of the time when a summary shows, versus 15% without one.
• Google does not ban AI content. It bans unhelpful content, however it was made. Real experience and trust are still the moat.
• Fix this first: feed the AI your own expertise and edit every line. Skip this: publishing raw AI drafts at volume.

To be honest, most of what you read about ai in seo is either breathless hype or quiet panic. Neither helps you decide what to do on a Tuesday morning with a real business to run.

So here’s the short version. AI changes how fast you can work, and it changes how Google shows your pages. It does not change what earns the ranking in the first place. The fundamentals that worked ten years ago still work now.

I’ve used these tools daily for a good while. Before that I grew a service business from 1 location to 9 over six years, on SEO and websites alone, through two big algorithm eras. This is the practical version: what AI is good at, where it quietly costs you, and the one knob to turn first.

What does AI in SEO actually mean?

AI in SEO means using machine-generated help for the research, drafting and analysis parts of SEO, while a human still owns the strategy and the final word. That’s the honest definition, stripped of the hype.

In practice it covers a few things that often get lumped together. There’s using AI tools (like large language models) to do SEO work faster. There’s optimising your site so AI search features show it. And there’s the worry about whether AI content gets you penalised.

Keep those apart and the topic gets a lot calmer. One is a workflow choice. One is a visibility question. One is a quality question. We’ll take them in turn.

Has AI changed how Google ranks pages?

No. The signals Google uses to rank pages are broadly the same ones it has used for years. AI features sit on top of the index; they don’t replace how the index is built.

Google has been clear about this in its optimisation guide: the best practices for AI features are the same best practices for ordinary search. In fact Google says a page needs no special optimisations to appear in AI features, just to be crawlable, indexable and genuinely useful.

This matches what I saw running an SEO-led business across several Google updates. The tactics at the edges shifted. The core never did: a site Google can read, content that answers the real question, and trust signals from the wider web. If someone sounds completely different in 2026 than they did in 2016, be careful.

What is AI in SEO genuinely good at?

AI in SEO is genuinely good at the slow, repeatable work: research, outlines, briefs, and tidying up your own drafts. Here’s what I’d actually use it for, today, without hesitation:

• First-draft research. Pulling together what a topic usually covers, so you start from a structure instead of a blank page.
• Outlines and briefs. Turning a keyword into a sensible heading structure you then rip apart and improve.
• Boring-but-needed text. Meta descriptions, image alt text, schema scaffolding, FAQ phrasing.
• Pattern-spotting. Clustering keywords, grouping search intent, summarising a long competitor page.
• Editing your own writing. Tightening a rambling paragraph you wrote, which keeps your voice but loses the waffle.

None of that is controversial, and the numbers show it. Around 86% of SEO professionals now use AI somewhere in their process. The tools have moved from novelty to normal.

It’s not just agencies, though. UK AI adoption is still uneven, with small firms reaching only 26% by 2025, so an owner who uses these tools well can still get ahead of slower local rivals.

That’s also why “I use AI” is no longer an advantage on its own. If everyone drafts with the same models, everyone’s first draft sounds the same. The advantage is what you add on top, which is the part AI can’t fake: your real experience.

Where AI in SEO quietly hurts you

The damage from AI in SEO is rarely a dramatic penalty. It’s slow, dull erosion: pages that read fine but say nothing only you could say.

The common failure pattern looks like this. An owner generates 40 posts in a weekend, publishes them all, and waits. Six months on, traffic is flat or down, because none of the pages earn a link, a citation, or a second visit.

There are three traps worth naming:

1. Volume over substance. More pages is not more rankings. Google’s helpful content approach is built to spot pages made for search engines rather than people.
2. No real experience. AI can’t tell a reader what went wrong on your last job. That first-hand detail is the thing that makes a page worth reading and worth linking to.
3. Confident nonsense. Models invent statistics and misstate facts. Every number and claim needs checking against a real source before it goes live.

The fix isn’t to avoid AI. It’s to treat its output as a rough draft from a fast junior, not a finished article from an expert.

How do AI Overviews change the game for small businesses?

AI Overviews mean fewer clicks from informational searches, so you should aim for fewer but more qualified visits rather than raw traffic. That shift is already measurable.

Pew Research found that when an AI summary appears, users click a normal result about 8% of the time, against 15% with no summary. Ahrefs put it more bluntly, measuring a 58% drop in clicks for the top organic result on queries with an overview. Search Engine Land’s coverage of the same study reached the same point: the summary answers the question, so the click never happens.

For a small local business, this matters less than you’d fear, and here’s why. People searching “what is local SEO” were never going to book you anyway. The searches that pay your bills are the ones with intent: “emergency plumber Slough”, “accountant near me”. Those still send clicks, because the searcher wants to contact someone, not read a paragraph.

Search Engine Journal’s advice for publishers adapting into 2026 lands in the same place: depth and originality beat volume. So put your effort where the booking happens, not where the curiosity lives. That usually means your local SEO and your service pages, not a pile of thin explainer posts that an overview will now eat.

Does Google penalise AI-generated content?

No. Google does not penalise content for being made with AI. It acts against content that’s unhelpful or made mainly to game rankings, whatever tool produced it.

Google’s own guidance says using automation to manipulate rankings breaks its spam policies, but that helpful content made with AI is fine. The test is the output, not the method. A useful page written with AI help passes; a useless page written by hand does not.

What actually protects you is E-E-A-T: experience, expertise, authoritativeness and trust. Of those, Google calls trust the most important. AI can help you write, but it can’t supply your years on the job or your name on the byline.

That’s the real reason the owner is the best SEO writer a business will ever have. The job of a good consultant, or a good AI workflow, is to structure and sharpen what’s already in the owner’s head, not to invent expertise from nothing. It’s the same model I use for SEO content on this very site.

The one knob to turn first

If you change one thing about how you use AI in SEO, make it this: stop asking AI to be the expert, and start using it to express yours.

The practical version is simple. Before you generate anything, write down the three things you know about this topic that a competitor wouldn’t: a mistake you see clients make, a number from a real job, a strong opinion. Feed those in. Then let the AI structure the piece around your input, not the other way round.

I learned the value of this the slow way. The business I grew didn’t win on clever tricks; it won because every location page and post carried real operating detail, built up patiently over years. Local SEO compounds, it doesn’t spike, and AI doesn’t change that maths. It just lets you produce the substance faster, if you actually have substance to produce. You can read my background if you want the longer story.

What does a safe AI-in-SEO workflow look like?

A safe AI-in-SEO workflow puts your own expertise in first, then uses AI to structure and speed up the rest. Here’s a process I’d be comfortable handing to any owner who wants the speed of AI without the risk:

1. Pick one keyword with intent. One page, one job. Commercial intent before vanity terms.
2. Brain-dump your own expertise first. Three to five points only you could write.
3. Let AI build the outline. Then cut, reorder, and add the sections it missed.
4. Draft with AI, section by section. Paste your expertise into each section as you go.
5. Edit hard, in your own voice. Remove anything generic. If a sentence could sit on any competitor’s site, cut it.
6. Fact-check every claim. Replace invented stats with real, cited sources.
7. Add the human layer. A specific example, a real number, a clear next step for the reader.

That last layer is everything. It’s also where most people stop too early, which is exactly why it’s still an advantage.

If you’re weighing whether to do this yourself or get help, the honest answer is it depends on your time, not your budget. Some owners enjoy the writing and just need the structure. Others would rather hand it over and stay the named expert. Both work, and both sit behind the SEO consulting here.

Where to start

AI in SEO is a multiplier, not a shortcut. It makes a good process faster and a bad process fail quicker. The owners who win with it are the ones who already had something worth saying.

If you want a read on where your own site stands before you add AI to the mix, the free audit is the place to start. Drop your URL in, get a score on the page, and a personalised report lands in your inbox a few minutes later. No call needed to get going, and you can take the report to anyone. Send it over and let me know what’s going on: free audit.