AI can do a lot of SEO. Whether it can do your SEO depends on which parts you actually need done. Because SEO is one area where AI has genuinely changed the job, and anyone telling you nothing has changed is selling you something. The honest answer isn’t yes or no, it’s a list of what it handles well and where it falls over, which is what this piece covers.
First, the thing nobody warned you about
The bigger story in SEO right now isn’t whether AI can write your blog posts. It’s that AI has changed what ranking is worth.
Google now shows an AI-written summary at the top of a lot of searches. Ahrefs analysed 300,000 keywords and found that when one of those summaries appears, the top result loses a large share of its clicks. Their first study put the drop at around 34.5%. When they re-ran it with later data, the gap had widened considerably.
Pew Research Center tracked the actual browsing of 900 people and found something similar from the user’s side. When an AI summary appeared, people clicked a normal search result in about 8% of visits. Without one, it was 15%. Almost twice as often.
So you can rank first and still get less traffic than you did two years ago.
This matters more than any question about which tool writes your content. If your SEO plan is “rank for lots of questions people ask,” a large chunk of that traffic now gets answered on the results page and never reaches you.
The good news is that the effect isn’t even. Question-style searches get hit hardest. Searches where someone is ready to buy get hit far less, because Google would rather show ads there. Which means the searches worth having are mostly still worth having.
What AI genuinely does well
Let’s give it credit where it’s due. There’s a lot of SEO work that used to eat hours and now doesn’t.
Keyword research at volume
Pulling thousands of search terms, grouping them by topic, spotting patterns. This used to be a spreadsheet slog. AI is fast and good at it.
First drafts
Give it a clear brief and it produces a serviceable draft in a minute. It won’t be your best page, but it beats a blank screen. We go into where that draft helps and where it hurts in our comparison of AI content tools and copywriters.
Technical checks
Finding broken links, missing page descriptions, slow pages, duplicate content. Repetitive and rules-based, which is exactly what software is for.
Rewriting for structure
Turning a wall of text into something with clear headings and a direct answer near the top. That’s worth doing now, because the AI summaries pull from pages that answer plainly rather than pages that ramble.
Spotting gaps against competitors
Feed it your pages and theirs and it will tell you what they cover that you don’t. Useful raw material, though it won’t tell you which gaps are worth filling.
Translation and local versions
Fast, cheap, mostly good.
If you or someone on your team knows SEO, these tools will make you meaningfully faster. That’s a real gain and we use them for exactly this.
Where it falls over
Deciding what’s worth ranking for
A tool will show you a keyword with high volume and low competition and mark it green. It doesn’t know that the people searching it will never buy from you. Half of the wasted SEO budget we see comes from businesses ranking successfully for terms that were never going to make them money.
Knowing what your customer is actually asking
Two people typing similar words can want completely different things. One is comparing suppliers. One is trying to fix a problem themselves. A page that serves the wrong one gets traffic and no enquiries.
Here’s a real example of the difference. Take a company selling payroll software. “How to calculate PAYE” looks brilliant on paper. Big volume, low competition, clearly related to the product. But the people searching it are mostly trying to work out one number for themselves this afternoon. They don’t want software. They want the answer, which Google now gives them at the top of the page anyway.
“Payroll software for small business South Africa” has a fraction of the volume. It also has people with budget and a problem. One of those pages pays for itself. A tool ranks them by volume and puts the wrong one first.
Anything involving other people. Getting mentioned, quoted, linked to, or reviewed still comes down to relationships and being worth talking about. No tool does that for you.
Being trusted. Google’s own guidance on helpful, people-first content leans heavily on signals that you’re a real business with genuine experience of the subject. Google has also been clear that it doesn’t ban AI-written content, but it does judge it on quality like anything else. Ten thin articles on a topic you have no track record in won’t get you there, and a site full of obviously generated filler can do active harm.
Knowing when to stop. The hardest call in SEO is admitting something isn’t working and moving the budget. Tools don’t volunteer that. They keep reporting.
The pattern
Notice what the “falls over” list has in common. Every item is a decision, not a task.
That’s the honest split. AI has taken over most of the doing. The value has moved to knowing what to do, and that gap has widened rather than closed, because now everyone can produce content quickly and the competition on volume is brutal.
Publishing more than your competitors used to be an edge. It isn’t anymore. Publishing the right things is.
The same pattern shows up in paid search, where Google’s automation now handles the bidding and matching but has no opinion on whether it’s chasing the right leads.
What to actually do about the AI summaries
Knowing your clicks are down isn’t much use on its own. Here’s what we’d change if this is affecting you.
Work out which of your pages are exposed
Not all of them are. Go through your top pages in Search Console and compare impressions with clicks over the past year. If impressions held steady but clicks fell, a summary has probably appeared above you. That’s the list to deal with first.
Move budget toward buying-intent pages
Service pages, comparison pages, pricing pages, location pages. These trigger AI summaries far less often, and the people reading them are closer to spending. If your content plan is mostly beginner explainer articles, it’s aimed at exactly the traffic that’s disappearing.
Answer plainly and early on the pages you keep
Summaries pull from pages that state things directly. A page that spends four paragraphs warming up before answering is less likely to be used. Lead with the answer, then explain it.
Stop counting traffic as the win
If your reporting is a traffic line going up, you’ll be misled in both directions this year. Count enquiries, calls, and revenue instead.
Build the things a summary can’t replace
Your own data, your own customer stories, your own opinions, your own tools and calculators. Anything that requires having done the work rather than having read about it.
So can you run your SEO on AI tools?
Yes, if you know SEO. The tools will make you faster and you should use them.
If you don’t, what usually happens is this. You publish steadily for six months. Traffic goes up a bit. Enquiries don’t move. Nobody can tell you why, because the reports show activity rather than answers.
Common questions
Will Google penalise us for using AI to write content? No. Google has said publicly that it judges content on quality rather than how it was made. What gets penalised is unhelpful, thin, or duplicated content, whoever or whatever produced it.
How long does SEO take to work now? Broadly the same as before, which is three to six months before you see meaningful movement on competitive terms, and longer on a new site. What has changed is that the payoff at the end is smaller for informational content and roughly unchanged for buying-intent content.
Should we stop writing blog posts? No, but change what they’re for. Posts that answer a simple factual question are worth less than they were. Posts that show your thinking, share your data, or help someone make a decision are worth more, because those are the ones that build trust and get referenced.
Is it worth optimising to get cited inside the AI summary? Being cited is better than not being cited. It is not a substitute for the click you used to get, because very few people click the citations. Treat it as brand visibility rather than traffic.
Can we just use an AI tool and check the output ourselves? That works if the person checking knows what good looks like. If nobody in the business can confidently say why one page should exist and another shouldn’t, the checking step isn’t really happening.
Where this leaves you
If you have SEO expertise in the building, use the tools hard. They will save you real hours and you’ll get better work done in the time they free up.
If you don’t, be honest that the tools are solving the easy half of the problem. The expensive mistakes in SEO have never been in the writing. They’re in ranking for things that were never going to make you money. If you’re weighing up whether to run this yourself or bring someone in, our full comparison of both routes is here: AI vs Marketing Agencies.
