Artificial intelligence SEO: how AI transforms search in 2025

Half the SEO industry spent 2025 panicking about AI killing search. The other half spent it selling "AI SEO" courses to the first half. We sat in the middle, running campaigns, watching the data, and quietly fixing things that broke.

So here is the honest version. Using AI for SEO is real, useful, and already baked into how we work at SEO Engico. It is also wildly oversold by people who have never had to explain a traffic drop to a client on a Friday afternoon. This piece separates the two, with numbers we have actually checked.

We are mechanical engineers by training, which means we treat marketing like a system. Find the bottleneck, fix it, measure the result. AI is just a new set of tools in that system. A genuinely good set. Not magic.

What does "AI for SEO" actually mean in 2025?

It splits into two completely different things, and conflating them is where most people go wrong.

The first is using AI as a tool to do SEO work faster: drafting content, clustering keywords, finding internal link opportunities, spotting technical issues, summarising competitor pages. This is the boring, profitable bit.

The second is optimising so AI engines mention you when they answer a question. That is when ChatGPT, Google's AI Overviews, Gemini or Perplexity synthesise an answer and decide which sources to cite. People call it GEO, AEO, AI search visibility, a dozen names. We mostly call it the thing clients are actually scared of.

Both matter. They need different playbooks. If a tool or an agency blurs them together, that is your first red flag.

The data on AI Overviews: how bad is the click problem, really?

Let's start with the headline that keeps founders awake. Yes, AI summaries are eating clicks. No, search is not dead. Both things are true at once, which is annoying but important.

The most rigorous study we have seen comes from the Pew Research Center, which analysed real browsing data from 900 US adults across 68,879 Google searches in March 2025. When an AI summary appeared, users clicked a traditional search result just 8% of the time, versus 15% when no summary showed. Links inside the summary itself? Clicked a measly 1% of the time. And people ended their browsing session 26% of the time after seeing a summary, compared with 16% without one.

Ahrefs went bigger and angrier. Their study of 300,000 keywords, comparing Google Search Console data from December 2023 against December 2025, found that the presence of an AI Overview correlates with a 58% lower click-through rate for the top-ranking page. That is up from 34.5% only eight months earlier. The trend is moving fast, and not in publishers' favour.

Here is our take, though. That 58% figure measures CTR on keywords that already trigger an Overview, not your whole site's traffic. Plenty of commercial and local queries never show one. The pages getting hammered are the thin informational ones that answered a simple question in 40 words. If that describes your content strategy, AI Overviews are not your problem. Your content was always replaceable.

How are UK users actually behaving?

Most stats on this topic are American. We work with UK businesses, so we care about the UK picture, and it is genuinely striking.

Ofcom, the UK communications regulator, published its Online Nation report in December 2025. ChatGPT pulled 1.8 billion UK visits in the first eight months of 2025, up from 368 million in the same period of 2024. That is roughly a five-fold jump in a year. Around 30% of searches now show an AI Overview, and 53% of adults say they see these summaries often.

Notice the framing in Ofcom's own data: most people are not seeking AI summaries out. They are just being served them. That tells you something the hype merchants miss. Behaviour is shifting because the interface changed, not because everyone suddenly fell in love with chatbots.

And Google has not collapsed. Statcounter puts Google at 91.42% of UK search market share as of April 2026, with Bing trailing on 5.93%. So when someone tells you to abandon Google and "optimise for ChatGPT" instead, they are advising you to ignore nine in ten of your potential customers. We do not recommend that. We recommend doing both, which is what our AI search visibility service is built around.

Traditional SEO vs GEO: is it really a different discipline?

This is where we are going to upset some people, including some who sell GEO packages.

Google published an official guide to optimising for its generative AI features, and the message is blunt. They state that the best practices for SEO continue to be relevant because their generative AI features are rooted in core Search ranking and quality systems. They go further: you do not need to create new machine-readable files, AI text files, or special schema.org markup to show up in AI search. From Google's perspective, optimising for generative AI search is optimising for the search experience, and is therefore still SEO.

Search Engine Journal covered Google's updated guide in May 2026 and summed it up neatly: Google's mythbusting section explicitly names tactics site owners do not need, including llms.txt files, content chunking and AI-specific rewriting. Some of those are the exact things GEO vendors charge a premium for.

Our honest position: GEO is not a new discipline. It is good SEO, plus a sharper focus on being a citable source rather than just a ranking URL. The fundamentals do not change. The emphasis does.

What does change is the goal. Classic SEO is zero-sum. One page wins position one, everyone else loses. AI answers can cite three, four, five sources at once, so the game becomes "be one of the trusted few" rather than "beat everyone to the top". That shift rewards genuine authority, original data and clear, well-structured pages. If you want the deeper version, we wrote a full breakdown of why most brands get GEO wrong that pairs nicely with this.

Where AI genuinely earns its keep in our workflow

Enough theory. Here is what we actually use AI for day to day, and where we have learned to keep it on a short leash.

1. Keyword clustering and intent mapping. Sorting thousands of keywords into topic groups used to take a day. AI does the first pass in minutes. We still check it, because models love to lump commercial and informational intent together, and that quietly wrecks your keyword strategy if you let it.
2. Content briefs. AI is excellent at pulling apart the top-ranking pages for a query and showing you what they all cover. We turn that into a brief, then a human writes the thing. Pages built entirely from AI tend to read like wallpaper, and Google's helpful-content systems notice.
3. Technical triage. Feeding crawl data to a model to flag patterns, broken internal links, orphaned pages, redirect chains, is fast and surprisingly good. It speeds up our technical SEO work without replacing the judgement of deciding what actually matters.
4. Spotting cannibalisation. AI is handy for finding pages competing for the same query, which has become more painful in the Overviews era. We covered the mechanics in our piece on keyword cannibalisation and AI Overviews.
5. First-draft outreach and reporting. Internally only. AI drafts the boring scaffolding, a human adds the bits that make it true and human.

The pattern across all of these: AI does the volume work, humans do the judgement. The moment you flip that ratio, quality falls off a cliff and you spend more time fixing output than you saved generating it. We learned that the expensive way.

The AI content trap nobody warns you about

Here is the thing that genuinely catches people out. Publishing a hundred AI-written articles feels like productivity. It is usually just future cleanup.

We have audited sites that mass-published AI content and watched their traffic flatline or worse. Not because Google detects "AI writing" and penalises it. Google has been clear it cares about quality, not the tool used. The problem is that fast, cheap content is almost always shallow content, and shallow content does not get cited by AI engines or ranked by Google. It just sits there.

If you are weighing this up, our take on whether Google actually cares about AI-detected content goes deeper. Short version: nobody is scanning for robots. They are measuring whether your page is the best answer. Most bulk AI content is not.

The content that does win in AI search is the opposite of mass-produced. Original research, first-hand experience, real numbers, a clear point of view. The exact things a language model cannot generate from thin air, because it has to be true. That is also why we keep publishing our own original research on AI search visibility rather than rehashing everyone else's.

How to actually get cited by AI engines

If your real worry is being mentioned inside ChatGPT or Google's AI answers, here is the practical sequence we follow. None of it requires a magic file.

• Answer the question early and plainly. AI engines lift clear, self-contained statements. Put a direct one or two sentence answer right under each heading, then expand. This helps featured snippets and AI citations at the same time.
• Build real authority and consistency. Make sure your brand facts match across your site, your profiles and third-party mentions. Models hate ambiguity about who you are. Our guide to getting your brand into AI answers walks through this properly.
• Publish things worth citing. Data, statistics, definitions, comparisons. If you are the source of a number, you become the link.
• Keep technical fundamentals clean. A page has to be indexable and eligible for a snippet to be eligible for AI features. Google said so directly. Boring, crucial.
• Track citations, not just rankings. Watch how often you show up across ChatGPT, Gemini and AI Overviews. If you want a head start, see how to get cited in ChatGPT and AI Overviews.

Notice what is not on that list: gaming the system. There is no shortcut. The brands getting cited are the ones who were already trustworthy and just made themselves easier to quote.

Should you build this in-house or hire it out?

Fair question, and the answer genuinely depends on your situation rather than what an agency wants to sell you.

Building in-house means hiring specialists, paying for enterprise tools, and budgeting months of learning time while the landscape keeps moving. The AI search rules in May 2026 are not the ones from January 2025. By the time an internal hire is up to speed, half of what they learned has shifted. For large organisations with hundreds of pages and constant optimisation needs, that investment makes sense.

For most small and mid-sized UK businesses, it does not. You are paying for a learning curve someone else has already climbed. That is the case for partnering: you get the frameworks, the tooling and the tracking from day one, without the recruitment risk or the gaps when someone leaves. Plenty of our clients run a hybrid, keeping an internal owner while we handle the specialist SEO and AI search work.

We are biased, obviously. But the bias is grounded in watching businesses burn six months and a salary trying to do this alone before calling us anyway. If you would rather skip that part, tell us what you are working on and we will give you a straight read on whether you even need help.

Our actual prediction for the rest of 2025 and into 2026

We will not pretend to know exactly where this lands. Anyone who does is guessing with confidence. But a few things look solid.

AI Overviews will keep expanding and clicks to thin informational content will keep falling. That is not reversible. The flip side: traffic that does come through will convert better, because the people clicking through after an AI summary already did their tyre-kicking. Fewer visits, warmer visits.

Google is not going anywhere in the UK at 91% market share, so "optimise for Google" remains the core job. AI engines are a growing second front, not a replacement. And the fundamentals, useful pages, real authority, clean technical structure, will matter more, not less, because AI systems are essentially a trust filter sitting on top of search.

The businesses that win the AI search shift are not the ones with the cleverest prompts. They are the ones who were already doing the work properly and adapted early. If that is the camp you want to be in, our team builds exactly this through our AI search visibility work, and we are happy to show you the data behind it before you commit to anything.