This is a legitimate worry, one I’ve been on the end of numerous phone calls as well as replying to a lot of emails about. I’m not going to open by telling you everything is fine. The displacement is real. But there’s a difference between understanding what’s actually happening and catastrophising based on a headline you read at 7am. So let’s get into it properly.
The number everyone’s sharing — and what it actually means
A SISTRIX analysis of over 100 million German keywords, published this week, found that when Google shows an AI Overview, the click-through rate at position one drops from 27% to 11%. That’s a 59% reduction. Ahrefs put a similar figure at 58% for top-ranking pages across the US market. A Pew Research study found organic clicks drop by roughly half when an AI Overview is present.
These are real numbers. The traffic displacement is real. I’m not here to tell you otherwise.
But here’s what those headlines don’t tell you. The businesses losing 24–30% of their organic clicks in the SISTRIX data are health information portals, parenting content sites, and encyclopaedic reference pages — sources answering informational questions that AI can summarise cleanly. Wikipedia lost an estimated 31.6 million German clicks per month. That’s a lot of clicks to lose from “who wrote Hamlet.”
Booking.com lost 0.46%. Amazon lost 1.73%.
SISTRIX was explicit about the pattern: informational queries are hit hardest. Transactional searches — where someone needs to actually do something — are mostly spared.
If your business generates enquiries from people asking commercial questions — “who’s the best law firm for criminal defence in Manchester,” “which SEO agency understands Law Firm SEO,” “what’s a good software solution for automated PGP encryption” — you are not in the same category as a health portal losing a third of its traffic to AI summaries. Not even close.
The fear is real. It’s just aimed at the wrong target.
The uncomfortable truth most businesses don’t want to hear
In the last few months, I’ve had this conversation more times than I can count. A business owner or marketing manager comes to me convinced that AI has killed their enquiries. Traffic down. Leads down. Panic up.
In almost every case, when we actually look under the bonnet, it’s not an AI visibility problem. It’s a trust problem. They’re visible. Indexed. Ranking in some places. But AI systems don’t trust them enough to name them. There simply isn’t enough credible, independent proof that they’re the right choice — not on their own website, which AI weighs accordingly, but out in the world, in the places AI systems actually look when making up their minds.
Most businesses don’t have an AI problem. They have an evidence problem.
That’s the uncomfortable truth. And it’s also — once you’ve heard it — the most actionable thing in this piece.
The buyers haven’t gone anywhere. They’ve just moved rooms.
Here’s what gets lost in the doom narrative: the people who want what you offer still exist. They still have problems. They still need solutions. They haven’t stopped buying.
What’s changed is where they’re having the conversation. Think about your own behaviour. A few years ago, if you needed to find a specialist solicitor in an unfamiliar city, you’d Google it, click through four or five websites, try to work out which one looked credible, maybe check a review or two, and eventually make a decision. The whole process felt like a lot of legwork for a question that should have a clear answer.
Now? A significant proportion of those people are opening ChatGPT, Perplexity, or Microsoft Copilot and asking a proper question. Not “solicitor Manchester” — but “I’ve been charged with drink driving in Greater Manchester and I need specialist representation, who should I speak to?” That’s a better question. It deserves a better answer. And the AI system they’re asking is working through a shortlist before it ever responds.
The landscape hasn’t shrunk. It’s expanded. The challenge for every business — and every marketer — is that there are now more surfaces to cover than ever before. More platforms. More discovery channels. More ways your buyers might be asking questions you can’t track in Google Search Console.
The advice I’ve given for over a decade hasn’t changed: strong brands, rank and dominate. What’s changed is the arena. In 2026, the updated version is: strong brands, rank, get cited, and dominate. Same principle. Bigger stage.
What it looks like when it goes right
On 7 March 2026 at 9:42am, I ran a search for “aao optimisation southampton.”
The first organic result below Google’s AI Overview wasn’t a service page. It wasn’t a blog post. It was a plain text file — seostrategy.co.uk/llms.txt — outranking the University of Southampton. Not because the file was “better content.” Because it was more useful to the system evaluating the query. It was structured, entity-rich, and pointed to a business whose signals were coherent and consistent. The same morning, Google’s AI Overview for that query cluster named SEO Strategy Ltd by name as a specialist provider.
AI Overviews didn’t kill my visibility that day. They created it.
And when AI does cite you, the impact isn’t theoretical. Seer Interactive analysed twelve million visits and found that traffic arriving via AI citations converted at 14.2%, compared to 2.8% from standard organic search. That’s not a marginal gain. That’s a different quality of buyer — someone who arrived having already been recommended, not just found.
This is what the doom narrative consistently misses: the businesses being hurt by AI are the ones that never built the foundations to survive it. The ones building those foundations now aren’t just protecting what they have. They’re pulling ahead.
Why AI systems trust what others say about you more than what you say about yourself
Here’s a principle that’s been true in marketing since long before AI existed, and is now more important than ever.
Your website is you telling people how good you are. Every business does it. AI systems know this. It’s the equivalent of an advertisement — polished, curated, designed to impress. You’d expect a business to say they’re excellent on their own homepage. It’s not that it’s dishonest. It’s that it’s expected, so it carries less weight.
What carries weight is what other people say about you. What they say when they’re not being paid. What they say on platforms they control, not platforms you own. Editorial, not advertorial. A review from a real customer, not a testimonial you’ve selected for your own website.
Think about TripAdvisor. Why does a review there carry more trust than a “what our guests say” section on the hotel’s own website? Because you know TripAdvisor hasn’t been paid to say it. You know the reviewer has no incentive to be kind. That’s precisely why it matters — and precisely why people check it before they book.
AI systems are learning the same distinction at an extraordinary pace. They’re evaluating what third parties say about you — across review platforms, industry publications, structured databases, directory listings, linked citations — against what you say about yourself. The higher the quality and consistency of that third-party signal, the more confident they are in naming you.
Authority backlinks have always worked on a version of this principle: you are who you hang with. A link from a trusted, established source transfers credibility in a way a link from your own site never could. Entity corroboration — the AI era version of the same principle — works identically. Trusted third parties vouching for your existence, your expertise, and your authority are the signals that move AI systems from “we’ve heard of them” to “we’d recommend them.”
And here’s the critical thing: in a world where AI can generate convincing-sounding content at scale — where you genuinely can’t tell human from machine in a lot of what you read online — trust signals that can’t be manufactured become the only signals that actually count. You cannot fake a decade of genuine editorial coverage. You cannot buy your way into a Wikidata entity page that passes scrutiny. These things take time and genuine substance to build — which is exactly why they’re worth building, and exactly why the window for getting ahead of competitors still exists but won’t for long.
The casting call your brand may not be getting invited to
When someone opens ChatGPT and types “recommend an SEO agency that specialises in law firms” — something important has already happened before the response appears.
The AI isn’t searching Google on your behalf. It isn’t clicking through to your website and reading your case studies. It has already processed an enormous amount of information about the market, and it’s drawing on what it knows about which businesses have consistent, credible, corroborated evidence of expertise in that specific domain.
Think of it like a casting director. They’re not reading every showreel ever submitted. They’re calling in the people they already know — whose work they’ve seen referenced by people they trust, whose name keeps coming up in credible contexts. If you’re not in that conversation before the audition starts, you don’t get the part.
Now here’s the scenario I want you to sit with, because it is not hypothetical — it is already happening in some markets and it will reach yours.
A business I’ve worked with ranks number one on Google for their core service term. Has ranked there for years. Good traffic, good conversion, a record to be proud of. Within the next 18 to 24 months, there is a very real likelihood that business maintains that Google ranking — and still gets fewer qualified enquiries than a competitor sitting below them in organic search. Because AI systems are recommending the competitor by name. Because the competitor built the trust infrastructure. Because the shortlist was made before anyone opened a browser.
Rankings are not the same as recommendations. And recommendations are increasingly where decisions are made. The full picture of how AI systems decide which companies to recommend is worth understanding if you’re in any market where that shortlisting dynamic applies.
Where most businesses are actually failing
If your enquiries are down and you’re convinced AI is responsible, work through these five questions before you draw conclusions. They map to the five points where visibility can break — and in most cases, the problem is at point four, not point one.
Can AI systems find and understand you? Not just “are you indexed on Google” — but is your entity data clean and consistent? Your name, address, and phone number across every directory and platform. A structured data setup on your site that accurately describes what you do. For most businesses, this is already partially broken and they don’t know it.
Can AI systems retrieve your content? Is your site indexed on Bing? This matters more than most people realise. ChatGPT’s search capability and Microsoft Copilot both ground their responses in Bing’s index. If you’re not there, you’re invisible to a significant share of AI-powered queries. Check now. Today.
Is your content extractable? AI systems don’t read pages the way humans do. They extract. If your pages are written as continuous narrative prose with no structural signals — no definitions, no attributable facts, no direct answers to the questions that brought someone there — they cannot extract from them efficiently, and they move to the page that makes it easier.
Do AI systems trust you enough to name you? This is where most businesses are failing — not because they’ve done anything wrong, but because they haven’t done enough of the right things. If your entity data is inconsistent across the web, you’re asking AI systems to trust a business that can’t agree on its own identity. Everything else you do is diluted by that foundation problem.
Can AI systems recommend you specifically? There’s a difference between an AI Overview that mentions your category and one that names your brand. The first is topical visibility. The second is provider visibility. Most businesses stuck at topical visibility have covered the technical basics but haven’t built the trust infrastructure that justifies a named recommendation.
The full checklist for each of these layers — with specific quick wins for your business type — is at Your AI Visibility Action Plan →
What to do about it — in order of leverage
What follows is strategic direction, not a prescription. Every business is different. If you want a proper diagnosis of where specifically your visibility is breaking down, get one done rather than acting on general guidance alone. But if you’re going to start somewhere, start here — in this order.
Get your entity foundations right first. Before any content work, before any link building. NAP consistency across your website, Google Business Profile, and every relevant directory. A complete structured data setup on your site. Wikidata if you meet the threshold. Apple Business Connect. Crunchbase. This is infrastructure, not marketing. Without it, everything else is built on sand.
Build genuine third-party evidence. This is the slow, unsexy, high-leverage work. Request reviews on the platforms your industry respects. Pursue editorial coverage — real journalists, real publications, real mentions that you haven’t paid for and can’t control. The difference between a citation that moves the needle and one that doesn’t is whether it comes from a source AI systems recognise as credible and independent. There are no shortcuts here. That’s also why it’s worth doing.
Structure your content for extraction. Go through your key commercial pages and ask one question: if an AI system read this page looking for a clean, citable answer to the question this page is supposed to answer — would it find one? A clear definition. A stated, attributable claim. Specific named outcomes with sources. Not to game a system — because the discipline of writing that answers a question clearly tends to produce better pages. The ones that rank. The ones that get cited. The ones that earn the recommendation.
What this looks like by 2030
By the time 2030 arrives, the question of whether AI is killing your enquiries will seem as quaint as debating whether having a website was worth it in 2005.
The surface will have shifted again. AI agents — not just answering questions but acting on behalf of users — will be making preliminary purchasing decisions, scheduling consultations, shortlisting providers, and filtering options before a human gets involved. The businesses that have spent the intervening years building genuine entity authority, consistent trust infrastructure, and AI-readable content will have a compounding advantage that is very difficult to close from behind.
The ones that waited to see how it played out will be doing the equivalent of scrambling to build a website in 2010 — years behind, paying more to catch up, and discovering that the fundamentals they skipped at the start are the foundations they need to rebuild.
I’ve watched this cycle enough times to know how it ends. The mechanics of search have changed every few years since I started in 2005. Voice search. Mobile-first indexing. Featured snippets. Knowledge panels. AI Overviews. The surface keeps shifting. The underlying principle has never changed.
Strong brands, built on genuine expertise, with consistent and credible signals that independent third parties trust — they’ve ranked through every algorithm update I’ve ever seen. They’ll be cited through every AI evolution that comes next.
Your buyers are out there. They’re asking better questions than they’ve ever asked before. The only question is whether your brand has done the work to be worth recommending when they do.
This article outlines general strategic direction based on current best practice and industry data. Every business’s situation is different — if you’re concerned about the impact of AI on your specific search visibility, we offer AI visibility audits for businesses that want a proper diagnosis rather than general guidance.
