Category: AI

There’s a quiet problem sitting under most of the AI SEO advice being sold to UK businesses right now. Almost all of it assumes your website is actually capable of doing what’s being asked of it. If you’ve been reading anything about AI search lately, you’ll have noticed most of it ends in the same place. The work you’ve already done counts; it just needs to catch up with how AI tools now consume it. Add the schema, write the FAQ blocks, and properly structure the content. 

For most UK SME websites, that’s a big assumption. And it’s the one thing nobody’s properly talking about. None of that work will catch up unless something more fundamental changes first. The site itself.

All of that work eventually depends on the website being fast enough and clean enough for AI tools to read and trust in the first place.

This is the part of the conversation that’s harder to have, because it goes a layer below content and SEO and lands in development territory. But it’s the layer that decides whether any of the AI work above it actually pays off. So let’s talk about it properly.

The Thing Nobody Mentions about AI Tools and Your Website

Most people picture AI search engines as patient. They sit, they read, they decide. That isn’t what happens. When ChatGPT, Gemini, AI Mode or Google’s AI Overview goes to retrieve a page, it isn’t browsing the way a human does. It’s running a tightly time-boxed fetch. 

AI bots behave very differently from Google’s traditional crawler. They prioritise speed and tend to be far less patient with slow-loading pages, which means a site that takes too long to respond often doesn’t make the cut at all. If your site takes too long to load, or if the HTML it returns is too heavy to parse cleanly, the bot gives up and moves on. It doesn’t reschedule. It doesn’t try again gently. It just doesn’t include you.

And here’s the bit that catches most business owners off guard. The page never appears in any kind of error log. The bot didn’t fail. It just chose a faster competitor. 

Your traffic dashboard looks fine. Your rankings look fine. You’re simply not being cited, and the only way you’d know is if you actually went and ran your customer queries through ChatGPT and Gemini yourself, which most businesses still don’t.

What we see consistently across the sites we audit is that strong performance doesn’t guarantee AI citation, but poor performance reliably prevents it. Speed isn’t a lever you pull to win. It’s the floor you have to stand on to be in the game at all. 

If your site sits below that floor, no amount of AI SEO or GEO advice will rescue it.

What “Too Slow” Actually Means In 2026

Google’s Core Web Vitals are the technical thresholds Google uses to measure whether a website performs well enough to be worth ranking. Most UK SME websites haven’t been audited against them in years, and the numbers worth knowing are:

• Largest Contentful Paint (LCP) is the time it takes for the main bit of content on your page to appear. The “good” threshold dropped from 2.5 seconds to 2.0 seconds. Anything slower starts to hurt you.
• Interaction to Next Paint (INP) replaced First Input Delay back in 2024. It measures how quickly your page responds to clicks, taps and key presses throughout the session, not just the first. Anything above 200 milliseconds is a problem.
• Cumulative Layout Shift (CLS) measures how much your page jumps around as it loads. Anything above 0.1 hurts your business.

Alongside those three, AI tools are paying close attention to two more things that don’t appear in the official Core Web Vitals at all. 

The first is Time to First Byte (TTFB), which is how long your server takes to send the first scrap of data after it’s been asked.  Anything above 200 milliseconds starts to push AI crawlers towards giving up. 

The second is HTML payload size. AI bots have memory and compute budgets. The leaner your raw HTML (before images or scripts), the easier it is for them to fetch, parse and use your content. Heavy HTML payloads make a site harder to read and quietly push you down the citation queue.

In our experience, we routinely audit UK websites that ship 2-3 MB of HTML per page. Those sites are quietly invisible to AI search, and no one is telling them.

You can check most of this for free in about ten minutes. PageSpeed Insights, Google Search Console, and Chrome’s built-in DevTools (right-click any page, select Inspect, then open the Lighthouse tab) all provide the relevant numbers.  Those ten minutes are the cheapest research you can do.

So, Why Are Most UK Websites Slow?

Two things, mostly and with time they compound. And neither of them is anyone’s fault exactly, which is why they’re so hard to spot.

The Page-Builder Problem

The vast majority of UK SME websites are built with visual page builders. WordPress with Elementor, Divi or WPBakery. Wix. Squarespace. GoDaddy’s site builder. Shopify with various theme builders sitting on top. These tools made building a website cheap, fast and accessible, which is genuinely a good thing. 

The trade-off, which nobody mentions when they’re selling you the build, is what they do to the underlying HTML.

Visual page-builders work by wrapping every element on your page in stacks of nested divs, so they can be moved, styled, and rearranged inside the editor. The result is HTML that’s three or four times longer than it needs to be, full of inline styles, and packed with JavaScript that runs on every page load. Add a few animations, a slider, a pop-up builder, a cookie banner, and you’ve got a page that takes 3 to 5 seconds to become interactive on a decent connection, and significantly longer on a slower one.

For a human user, that’s frustrating but tolerable. They wait. For an AI crawler operating on a one-second budget, it’s a polite refusal. They leave.

This isn’t an argument against WordPress or against page-builders themselves. It’s an argument against using them lazily. A WordPress site built with disciplined development hours, sensible plugin choices, and proper performance tuning can be fast and AI-readable. A WordPress site built by dragging and dropping in Elementor and ticking every “add this feature” box can’t.

The Plugin Tax

The other half of the problem is plugin sprawl. The average UK WordPress site we audit runs somewhere between 25 and 40 active plugins. 

Each one was installed for a reasonable reason at the time. SEO plugin. Form plugin. Backup plugin. Security plugin. Booking plugin. Live chat. Reviews. Cookie compliance. Newsletter signup. Image gallery. Social sharing. Accessibility widget. Page builder. Page-builder add-on. Page-builder add-on extension.

Each plugin loads JavaScript, often loads its own CSS, often makes its own database queries and often loads on every single page, whether it’s needed there or not. The cumulative effect is brutal. The HTML balloons make the page slow to interact with, and the page’s structure becomes harder for an AI crawler to make sense of, even when it does manage to download it. 

The actual content is buried under layers of widget markup.

You can have the world’s best-written service-area page with perfect FAQ blocks and beautiful schema, and an AI tool will still skip it if the page takes 4 seconds to load and the actual answer is hidden 1.4MB deep in the HTML.

What “Fast And AI-Readable” Actually Looks Like

Here’s the bit most UK businesses haven’t been told. Sites that AI tools love tend to share a few characteristics, and almost none of them have to do with how the site looks.

• Clean, semantic HTML – Headings that are actually <h1> and <h2> tags, not styled divs. Lists that are actually <ul> and <li>. Article content wrapped in <article>. The structure of the document tells the AI tool what’s important, before it even reads the words.
• Server-rendered or static HTML – The actual content of the page is in the initial HTML response, not built in the browser by JavaScript afterwards. AI crawlers vary in how well they handle JavaScript rendering, and the safe assumption is that if your content isn’t there in the initial HTML, some of them won’t see it at all.
• A small, lean payload – Modest plugin footprint. Compressed images. JavaScript only where it’s actually needed. CSS that doesn’t ship the styles for every page on the site every time.
• Quick server response – Decent hosting (not the £3-a-month shared plan), proper caching, and a CDN if you’re serving any sort of geographic spread.
• Schema in the right place, in the right shape – Not seven different plugins all generating overlapping schema, which is what we see most of the time.

None of this requires throwing your website away. Some of it does require having a developer (an actual one, not a designer with a page builder) take a proper look at the foundations.

What You Can Do Without Rebuilding

Most UK businesses can’t credibly rebuild their website tomorrow, and most don’t need to. There’s quite a lot you can do on the existing site to get meaningful performance back.

The first thing is a plugin audit.

Open your WordPress admin, look at every plugin, and ask honestly: do we actually use this? We routinely deactivate 30 to 50% of plugins on first audits, and the site immediately gets faster and tidier without losing any functionality that the business actually noticed.

The second is image optimisation. 

Most UK SME sites are still serving uncompressed JPGs and PNGs at desktop size to mobile users. Modern formats like WebP and AVIF significantly reduce image file size. There are plugins that do this automatically, and yes, that’s one plugin worth keeping.

The third is to defer or remove non-critical JavaScript. 

Things like chat widgets, pop-up builders, and tracking scripts often load and run before the main content does. Pushing them to load later, or only on the pages that actually need them, often shaves a full second off LCP on its own.

The fourth is to look at your hosting honestly. 

If you’re paying £4 a month for shared hosting and wondering why your site is slow, you’ve found your answer. Decent UK hosting costs £20 to £50 a month, depending on the size of the site, and the difference in TTFB is often the single biggest performance gain you can buy.

The fifth is to clean up your schema. 

Pick one source of truth (usually a single SEO plugin properly configured), and turn off the schema output on every other plugin that’s quietly generating its own. Duplicated and conflicting schemas are a real problem on UK sites and a real reason AI tools struggle to parse them.

Together, those five fixes move most UK SME sites from being skipped by AI bots to being properly read by them. That’s the threshold that matters.

When a Rebuild Is Genuinely The Right Call

Sometimes the foundation is past saving. A few honest signs:

• The site is built on a page builder that’s been replaced or abandoned by its developer (older WPBakery setups, legacy Visual Composer builds, certain GoDaddy generations).
• LCP and INP are bad and can’t be fixed without restructuring the whole template, because the slowness is baked into the theme.
• The site is genuinely old (5 to 8 years old) and has accumulated so many half-implemented changes that no one on your team can confidently say what’s safe to remove.
• You’ve already spent meaningful money on plugins, optimisation, and “speed services”, and the numbers haven’t moved.

If two or three of those sound familiar, you’re probably looking at a rebuild. The fear most UK businesses have is that a rebuild will tank their rankings, and that fear is reasonable, because most rebuilds are handled badly. 

A proper migration, planned by people who understand both SEO and development, doesn’t have to cost you rankings. We’ve done dozens of rebuilds where the new site outranks the old one within weeks because the foundations are now actually capable of supporting the content. Anyway, that’s a different conversation, and one we’d rather have once you’ve seen the audit numbers for yourself.

How Does This Connect Back To Your AI Search Visibility

Here’s the part that matters. All the AI SEO advice floating around assumes the website itself is something AI tools can actually read.

If your foundations are sound, all that work compounds beautifully. Schema gets parsed correctly. FAQ blocks get cited. Service-area pages get recommended in AI answers. Reviews get pulled into AI summaries. The whole investment starts paying off in places it never used to.

If your foundations aren’t sound, none of it quite works. The AI tool gives up before it sees the schema. The FAQ block lives at the bottom of a 3MB page and never gets read. The service-area page is technically beautiful, but it never makes the cut because INP is at 800 milliseconds, and the bot moved on. You end up paying for the GEO work, not seeing the return, and concluding that AI search is a fad, when actually the website was the problem all along.

Which is why we’ve started running performance and AI-readability audits as the first step in any new client engagement. Not because we love technical work for its own sake, but because the technical layer turns out to be the difference between AI search advice that pays off and AI search advice that quietly doesn’t.

What To Do Next

If you’ve read this and you’re not sure whether your website is fast enough or clean enough for AI tools to read properly, the cheapest first step is to run your homepage and your top three commercial pages through PageSpeed Insights. 

If LCP is above 2 seconds, INP is above 200 milliseconds, or your HTML payload is over 1MB, you’ve got something to look at.

If the numbers are bad, or you’d rather have someone else look properly, book a free website performance and AI-readability audit with us. We’ll show you exactly where your site sits against the new 2026 thresholds. 

We’ll show you where AI tools are giving up on your site, what’s pulling your performance down, and what it would take to fix. No assumption a rebuild is the answer. Just a straight read of where you stand.

The thing worth remembering is this. AI search rewards businesses that have done the technical groundwork. That’s actually good news for the kind of UK businesses that take their websites seriously, because it puts the bar back where it should be: on whether the work is done properly, not on who can shout the loudest about doing it. 

The businesses that quietly fix their foundations now will be the ones recommended by ChatGPT, Gemini, AI Mode, and AI Overview, while their competitors are still wondering why their traffic looks healthy, but their enquiries are quietly drying up.

Remember, Foundations first. Everything else follows.

Let’s stop dancing around the elephant in the room. If you open LinkedIn today, you’ll see two types of lies. Type A: “AI is just a tool and everything will stay the same!” Type B: “The robot apocalypse is here, and we’re all going to be living in pods by 2027.”

Both are wrong. But here is the uncomfortable truth that most AI consultancies won’t tell you: Yes, AI will replace people. Specifically, it will replace the people in your business who spend four hours a day acting as a human bridge between two software packages.

At Trident, we see it every day. We call it the “Copy and Paste Tax.” And if you’re paying it, you’re not just losing money; you’re suffocating your most expensive, most talented assets.

The Hidden Killer: The Copy and Paste Tax

Think about your most senior, most creative staff member. Now, watch them work. How much of their day is spent moving data from an email into a CRM? From a spreadsheet into a proposal? From a Zoom transcript into a project management tool?

That is the Copy and Paste Tax. It’s the friction of modern business. We’ve been sold a lie that “digital transformation” means having a dozen different SaaS tools. In reality, we’ve just created a dozen digital silos, and we’re using human beings as the glue to stick them together.

The Question You Should Be Asking: Did you hire an expert to solve client problems, or did you hire an overpaid data entry clerk with a fancy job title?

When you lean into outsourced AI and proper workflow automation, you aren’t just “speeding things up.” You are collapsing the distance between the data and the decision. The goal isn’t to move tools; it’s to make the tools talk to each other so your staff don’t have to.

From “Process Monkeys” to Rainmakers

Here is the slightly controversial bit: Most of your staff are currently “Process Monkeys.” They spend 60% of their brainpower following a manual, 30% on internal politics, and maybe 10% on the actual genius you hired them for.

If AI replaces “half your workforce,” it should be the half of their to-do list that is beneath them.

Imagine your Account Manager. Right now, they spend Monday morning updating status reports. By the time they actually call a client on Tuesday afternoon, they’re drained. They’re “working,” but they aren’t generating revenue.

What if they never had to touch a status report again?

What if, instead of 10 clients, they could handle 40, because the AI handles the onboarding, the scheduling, the follow-ups, and the data entry? Suddenly, that staff member isn’t a cost center; they are a Rainmaker.

The “High-Touch” Paradox

Firms adopting AI aren’t shrinking; they are becoming “High-Touch.” By automating the “Low-Touch” (emails, scheduling, data syncing), you free up your team to do the one thing AI cannot do: Build a human relationship.

In a world where every business is using AI to send generic emails, the firm that sends its best people to sit across a table (or a focused video call) from a client is the one that wins. AI is the engine room that allows your staff to be on the bridge of the ship.

Why Efficiency is a Defensive Trap

Most CEOs look at AI and think, “If AI makes my team 50% more efficient, I can fire 50% of my team.”

That is “Small Business Thinking.” It’s defensive. It’s a race to the bottom.

The “Growth Mindset” version? “If AI handles 50% of the administrative burden, my team can spend 100% more time in front of my clients.”

The firms we work with that are adopting AI at a foundational level aren’t shrinking—they’re scaling. They keep their headcount (or even grow it) because they’ve suddenly unlocked the capacity to handle double the client load without a dip in quality. They aren’t firing their staff; they’re finally deploying them.

Opinion: If your business model relies on charging clients for “hours worked” on administrative tasks that a GPT agent can do in six seconds, your business is already dead. You just haven’t realised it yet. Value is moving away from process and toward partnership.

The Trident Approach: Connect, Don’t Move

One of the biggest fears in AI implementation is the “rip and replace” nightmare. Business owners think they have to migrate their entire infrastructure to some new “AI-first” platform.

Rubbish.

The real magic happens when you connect your existing workflow. You stay in the tools you love—Slack, Outlook, Monday.com, Salesforce—but you remove the manual steps.

• Before: A lead comes in → Staff member reads email → Checks calendar → Manually types reply → Updates CRM → Sets a reminder.
• After: AI parses the lead → Checks availability → Drafts a personalised response based on previous client history → Updates the CRM → Notifies the staff member only when the meeting is booked.

The staff member didn’t do “work.” They did “business.” See the difference?

Putting Your Key Assets Where They Belong

Your staff are your key assets. But right now, they are probably hidden behind a wall of browser tabs. Every minute they spend “processing” is a minute they aren’t:

1. Building a deeper relationship with a high-value client.
2. Thinking of a creative solution to a complex problem.
3. Spotting the next £100k opportunity.

AI handles the “What” and the “How.” Your people handle the “Who” and the “Why.”

Is It Time to Outsource Your AI Strategy?

You have two choices. You can try to figure this out internally, letting your IT guy play with prompts in his spare time. Or, you can treat AI like the critical infrastructure it is.

As an AI consultancy, Trident doesn’t just give you a list of prompts. We look at your business’s plumbing. We find where the “Copy and Paste Tax” is highest, and we eliminate it, shifting your staff from the back office to the front line.

So, is AI going to replace half your workforce?

Only the half that was acting like a robot anyway. And frankly, your best people will thank you for it when they finally get to do the job you hired them for.

Stop paying the Copy and Paste Tax.

Would you like me to draft a “Staff Value Map” to help you identify which 50% of your team’s current tasks should be handed to AI tomorrow?

Ready to Stop Paying the Copy and Paste Tax?

If this article has hit a nerve, it’s probably because you already know where the inefficiencies are hiding in your business. The next step is doing something about them. At Trident, we work with ambitious businesses to identify exactly where AI can free your people, sharpen your processes, and accelerate your growth without ripping out the systems you already rely on. We’re based at The Silk Warehouse, Druid Street, Hinckley, LE10 1QH, and we’d love to have that conversation. Give us a call on 01455 557766 or visit wearetrident.co.uk to book a call with the team. The robots aren’t coming for your best people, but your competitors might be, if you don’t act first.

If you’ve spent the last year reading marketing blogs, you’d be forgiven for thinking that traditional SEO is on life support and that you need to start again from scratch. Almost every article we read pushes the same message:  AI search has changed the rules, you’re behind and unless you act now, you’ll be invisible. 

The urgency is real, but the conclusion is wrong. 

From where we stand, working alongside UK businesses every week, the reality is far more reassuring. Generative engine optimisation, or GEO, isn’t a replacement for SEO. It’s an extension of it. The websites cited by ChatGPT, Gemini, Google’s AI Mode and AI Overviews are, in most cases, the same websites that were already doing the SEO basics properly. The work you’ve already invested in your site is not wasted. It’s the foundation. What’s needed now is a few additional layers on top, and that’s where this guide comes in.

Why Is Everyone Suddenly Talking About  AI Citations?

Because the way people search has shifted, and the click economy is shifting with it. AI Overviews now appear in a significant share of Google search results (estimates range from 16% to 50%, depending on the study and category), and click-through rates on those queries have dropped meaningfully. One large study of over 300,000 keywords found a 34.5% drop in CTR for top-ranking pages when an AI Overview appeared above them.

Pages that get cited inside the AI-generated answer earn meaningfully more clicks than those that don’t. The position you used to fight for – page one, position three or four still matters, but it isn’t the prize on its own anymore. The prize is being the source of the  AI quotes.

The same shift is happening across every major  AI search surface. ChatGPT browses the web through Bing’s index and surfaces a small handful of cited pages. Gemini powers Google’s  AI Overviews and  AI Mode, drawing from Google’s own index. Each of these engines is making editorial decisions about which businesses to mention by name and link to and if your name isn’t in the answer, you don’t exist for that user.

So the question for any UK business owner or marketing manager is no longer just am I ranking? Am I being cited?

How Do  AI Engines Actually Choose What To Cite?

Let’s understand this part in simple terms. When you ask an  AI a question, the engine doesn’t just generate an answer from memory. It searches the live web (or its recent index), retrieves a set of pages it thinks are relevant, reads them, and then decides which ones to quote and link in its response. The technical name for this process is retrieval-augmented generation, or RAG.

Three things determine whether your page makes the cut:

• Can the engine find your page in the first place? – This is where traditional SEO sits: crawlability, indexation, internal linking, page speed, mobile responsiveness, and ranking position. If your page can’t be found, nothing else matters. This is the floor.
• Can the engine extract a clean answer from your page? –  AI engines aren’t trying to read your whole article. They’re looking for a specific, self-contained passage that answers the user’s question. If your answer is buried in paragraph six, behind three paragraphs of context, the engine will skip you and pull from a page where the answer appears in paragraph one.
• Does the engine trust you? Authority signals matter enormously. Brand mentions across the web, third-party reviews, citations on credible publications, consistent business information, and a clear sense of who you are as an entity all feed into whether the model treats you as a source worth quoting.

Notice that the first of those three things is straight, conventional SEO. Your existing investment isn’t redundant; it’s the entry ticket. Without it, the rest doesn’t apply. With it, a few targeted changes can multiply what you’re already getting.

What Do All Four AI Search Surfaces Reward In Common?

Before we look at how ChatGPT, Gemini, AI Mode, and AI Overview behave differently, it’s worth understanding what they share. From our work with UK clients across legal, healthcare, e-commerce and B2B services, the pattern is consistent.

All Four Reward Semantic Completeness. 

Semantic completeness sits at the top of the list. We’re talking about pages that fully answer a question without requiring the reader (or the AI) to click elsewhere. Research from across the GEO industry shows that content scoring high on semantic completeness is around four times more likely to be cited than content that only partially addresses a topic.

A page on “VAT registration thresholds” that also explains who needs to register, when to register, what happens if you miss the deadline, and how to deregister will outperform a page that only defines the threshold itself.

Clear, extractable answers near the top of the page are the next thing  AI engines look for. Roughly 44% of all  AI citations come from the first third of a page. If your most important sentence is sitting in your conclusion, you’ve buried it.

Structured data matters too. FAQ Page, Article, How To, Organisation and Local Business schema markup don’t directly cause citations, but they help  AI engines parse your content accurately. Pages with proper schema get classified correctly, and that’s what makes them eligible for citation in the first place.

Topical authority is recognised and credited heavily, which means clusters of interlinked content rather than isolated posts. 

A single excellent blog post on “commercial property surveys” is less citable than a hub of fifteen interconnected articles covering surveys, valuations, leases, dilapidations and SDLT.

And finally, freshness. Content meaningfully updated within the last 30 to 90 days is far more likely to appear in AI responses than content last touched two years ago. 

The word “meaningfully” is doing a lot of work in that sentence. And yes, before you ask, quietly changing the publication date from 2024 to 2026 and calling it a refresh doesn’t count. The  AI knows. We know. Your readers know.

How Does ChatGPT Decide Who To Cite?

ChatGPT, when in browse mode, retrieves results through Bing’s index. This is the single most overlooked technical fact in the  AI SEO conversation. If you’re not indexed in Bing Webmaster Tools, you’re invisible to the fastest-growing  AI search channel in the country.

We routinely audit UK client sites that rank well on Google but have never been submitted to Bing. Fixing this takes about twenty minutes and is the single highest-leverage action most businesses can take in their first week of GEO work.

Beyond that, ChatGPT favours a few specific things:

• Confident, definite language. The claim “X reduces costs by 20%” is cited. “X may potentially help reduce costs” gets skipped. Hedged writing is invisible to the model.
• Clear question-and-answer structures. Real questions as headings, direct answers underneath, no warm-up paragraph in between.
• Recognised brand authority across third-party platforms. Reviews on Trustpilot, Google, Yell, Feefo and trade-specific platforms all feed into ChatGPT’s sense of whether your business is real and credible.

How Does Gemini Decide Who To Cite?

Gemini draws from Google’s index, which means strong traditional Google SEO is already a meaningful advantage. On top of that, Gemini leans heavily on a few specific signals:

• E-E-A-T credentials. Experience, Expertise, Authoritativeness and Trustworthiness, the four signals Google has been talking about for years, now carry even more weight. Author bios, qualifications, real case studies and demonstrable first-hand experience matter more than ever.
• Properly structured author and team information. The credentials you already promote on your “About” and “Team” pages are exactly what Gemini is looking for, provided they’re marked up with Person schema rather than left as plain text.
• Verifiable third-party links. Connecting your team profiles to LinkedIn, regulatory bodies (SRA, RICS, ICAEW, GDC, FCA), and professional associations gives Gemini the external validation it needs to trust the credentials you’re claiming.

For UK service businesses like solicitors, surveyors, accountants, dentists, and financial advisers, this is genuinely good news. The hard work of building credibility has already been done. The job now is making sure that credibility is structured in a way that Gemini can actually read.

How Does Google  AI Mode Decide Who To Cite?

 AI Mode is Google’s conversational, multi-turn search experience. The one that lets users ask follow-up questions and have a proper back-and-forth rather than firing off single keywords. Under the bonnet, it uses a process called query fan-out:

• The system breaks the original question into sub-questions. – A query like “do I need a solicitor for a boundary dispute” might internally fan out into “what counts as a boundary dispute,” “can I resolve a boundary dispute without a solicitor,” “what does mediation cost,” “what happens if it goes to court,” and several more.
• It retrieves answers for each sub-question separately – Different pages, sometimes different domains, for each one.
• It synthesises everything into a single response – a single AI Mode answer might pull from seven or eight different sources, stitched together.

What this means in practice is that you don’t have to be the best page for the exact query the user typed. You have to be a strong page for one of the sub-questions the engine generates internally. The more sub-questions you cover well within a content cluster, the more entry points you create and the more times your business gets quoted across the same conversation.

We see this most clearly in regulated sectors. A law firm with one strong page on “boundary disputes” will struggle to be cited consistently. A law firm with twelve interconnected pages covering boundary disputes, party walls, easements, restrictive covenants, adverse possession, and the relevant case law will be cited across dozens of related queries, often multiple times within the same AI Mode conversation.

This is the single biggest argument for content clusters in 2026. One excellent page is no longer enough. The work is in building the surrounding ecosystem.

How Does  AI Overview Decide Who To Cite?

 AI Overviews are the AI-generated summaries you now see sitting at the top of standard Google search results, powered by a custom Gemini model. Not every search triggers one, and understanding when they appear is half the battle:

• Informational and how-to queries are the main triggers: “How does,” “What is,” “Why does”, and “Can I” questions almost always generate an  AI Overview.
• Longer queries are far more likely to trigger one: Searches of eight words or more are several times more likely to produce an  AI Overview than short keyword searches.
• Navigational and transactional queries usually don’t trigger one: “Tesco login” or “buy running shoes Leicester” tend to skip the  AI Overview entirely.

There are other studies with similar conclusions as well. If you want a better understanding, we can rewrite this section to:

• A recent Ahrefs study found that only around 38% of pages cited in AI Overviews now also rank in Google’s top ten organic results, down from 76% in their previous study. Other studies put the figure lower again. Either way, the direction is the same: AI Overview citation is decoupling from organic ranking.
• AI Overview citation is decoupling from organic ranking. A well-structured, semantically complete page sitting at position 14 can now be cited above a thinner page at position 3.
• Structure and answer quality are starting to matter as much as authority. The old assumption that ranking high automatically protects your visibility no longer holds.

For UK SMEs, this is a genuine opportunity rather than another threat. You don’t necessarily need to outrank a national competitor to be cited alongside them in an  AI Overview. 

You need to provide a cleaner, more complete answer than they do, and that’s a much more achievable target for a smaller business with sharper, sector-specific content than a sprawling enterprise site with thin coverage of everything.

In short, the playing field has tilted slightly in favour of smaller, more focused operators for the first time in years. Worth taking advantage of while it lasts.

How To Prepare My Website for AI-Driven Search?

If your SEO basics are in good shape, the work ahead is honestly very manageable. 

What Changes Should You Make First?

If we were briefing a UK client on a first 30-day GEO sprint, this is the order we’d work in:

1. Build an SEO and  AI SEO strategy and a Content Brief: Shift the goal from chasing top-10 SERP links to earning  AI Overview citations. Cover keyword research, site structure, landing pages, Google Business Profile and a full audit. Create a well-researched industry-standard content brief with semantic keywords and contextual content that aligns writers, SEO, and website design around the same outcome.
2. Run a site audit with robots.txt and a customised llm.txt setup: Confirm that GPTBot, CCBot, Google-Extended and ClaudeBot aren’t accidentally blocked, and set clear  AI crawl rules. Core SEO signals like crawlability, indexation, page speed, and mobile usability remain non-negotiable. They’re the prerequisite for  AI eligibility, not an alternative.
3. Standardise citations across the web:  Align name, address and phone number across UK directories, review platforms and local listings. Inconsistent NAP data quietly undermines both local rankings and  AI trust signals. Most businesses don’t realise how much of it is wrong until someone actually looks.
4. Start knowledge-graph-driven blog rewrites: Refresh top-performing pages by mapping entities, authoritative sources and the relationships between them. (Not by editing the publish date and calling it done.)
5. Set up SEO and  AI visibility tracking. Implement SEO across Google, AI Overviews, AI Mode, ChatGPT, and Gemini. Continuous monitoring of site health, rankings and  AI prompts. This shows where you’re being cited, where you’re not, and who’s quietly taking the space that should be yours.

That’s the first 30 days. None of it dismantles your existing SEO; all of it strengthens it. Our 180-day plan takes it full circle from there. 

Want to know more? 

Book a free  AI SEO audit. 

We’ll show you where your brand stands across Google AI, ChatGPT and Gemini, and what it’ll take to close the gap.

What Can You Stop Worrying About?

Honestly, most of it.

Backlinks still matter. Long-form content still matters. Local SEO matters more than ever, because AI Overviews lean heavily on Google Business Profile data, NAP consistency and local citations. Page speed, mobile responsiveness, and well-written content are all still doing their job. Anyone telling you traditional SEO is dead is selling you something.

What’s changed is the bar. The businesses winning in 2026 are those doing SEO fundamentals properly and layering GEO on top.

Where UK businesses go wrong is at the two extremes. Panic-buying a “GEO-first” pitch, or freezing entirely because the topic feels overwhelming. Both end up in the same place: slipping quietly down the AI rankings while competitors get cited.

The middle path needs people who understand both the SEO craft and the newer mechanics of how LLMs retrieve and cite content. That combination is still hard to find in-house, and it’s a big reason our clients reached out to us in the first place. 

It really does matter. But it doesn’t mean starting again. It means doing the next bit properly, and that’s the bit we’re good at.