AEO: what the BBC got right and what they missed

HubSpot lost 140 million website visits last year, and when the BBC reported it a week ago the marketing world acted shocked.

We weren’t, because our clients started seeing this shift 18 months back.

One local services client in the West Midlands lost 34% of their informational traffic between March and September 2025. But their AI-referred visitors, the ones arriving via ChatGPT and Google AI Overviews, were converting at nearly five times the rate of organic. The traffic dropped while the revenue climbed, and that pattern kept repeating across different client sites in completely different industries.

The BBC spoke to HubSpot’s CMO, a spice company, and a builders’ merchant, and all three confirmed what we’ve been telling clients since late 2024. Businesses that don’t structure content for AI answer engines are watching their traffic vanish.

But the article barely scratched the surface. Here’s what they got right, what they missed, and what actually needs to happen.

What is answer engine optimisation?

AEO is the practice of structuring website content so AI tools find it, understand it, and cite it when answering user questions.

Those AI tools include ChatGPT, Google AI Overviews, Perplexity, Claude, and Gemini. Each one works differently with its own citation preferences. We’ll get to those.

Traditional SEO targets search engine rankings while AEO targets citation in AI-generated answers. You need both running in parallel. If you’re only doing SEO in 2026, you’re building half a house and wondering why the rain still gets in.

The BBC used the term “answer engine optimisation” and mentioned “generative engine optimisation” (GEO) as an alternative. Both describe the same discipline. AEO is the name gaining traction in the UK.

Why are businesses losing website traffic to AI?

Google AI Overviews now trigger on 25% to 48% of queries depending on methodology, and click-through rates drop 60% to 70% when they appear.

The numbers are getting worse every month. BrightEdge tracked AI Overview appearance rates across industries from February 2025 to February 2026 and the expansion is staggering. Restaurants went from 10% to 78%. Education from 18% to 83%. Healthcare from 72% to 88%. The businesses that haven’t started adapting are running out of runway.

Bodnar from HubSpot confirmed that click-through drop in the BBC piece, and it lines up precisely with what we track across client sites using Search Console filtered by queries where AI Overviews fire. The before-and-after on those specific queries is brutal. In AI Mode specifically, Seer Interactive found 93% of sessions end without a single click to an external website.

Here’s the part most people miss though. The “AI search is tiny” narrative is dead. It was always based on incomplete data that only measured web traffic. Graphite’s March 2026 study combined web and mobile app usage across all major AI platforms and found they generate 45 billion monthly sessions globally. That’s 56% the size of all search engine volume combined. The reason everyone underestimated it is that 83% of AI usage happens inside mobile apps. Web analytics tools miss the real picture by 4 to 5 times.

Google still handles the majority of search activity. Their share dropped from 89% in 2023 to 71% in Q4 2025 according to Graphite. But total search volume actually grew 26% globally when you combine traditional and AI search. The pie got bigger. AI search is additive right now, not cannibalistic.

But the people using ChatGPT for research tend to be your highest-intent buyers.

MKM Building Supplies said exactly this in the BBC article. Their AI-referred visitors buy at significantly higher rates than organic visitors. AuthorityTech and First Page Sage data from early 2026 puts ChatGPT-referred traffic conversion at 11.4% versus 5.3% for organic. More than double the conversion rate. A physiotherapy clinic in Halesowen had service pages for each treatment type that ranked reasonably well but got zero AI citations. We built a content cluster around sports injury recovery covering rehab timelines for specific injuries, when to use ice versus heat, return-to-training protocols for runners, and what to expect from your first physio appointment. Within three months, the “first appointment” guide was being cited by ChatGPT in four out of ten test prompts for “do I need a physio or GP for back pain.” The service pages never got cited once. The educational content surrounding them did.

What did the BBC get right?

The BBC’s reporting on HubSpot, Spice Kitchen, and MKM Building Supplies nailed three points that matter for any business adapting to AI search: content chunking, topical authority, and E-E-A-T as baseline citation requirements.

Three things stood out from their piece.

Content chunking works. HubSpot broke their long product articles into smaller, self-contained chunks that AI tools can extract and cite individually. AI answer engines pull content paragraph by paragraph. A 3,000-word wall of text with no clear heading structure gets passed over for a competitor who made the same information easier to find.

Topical authority beats product pages. Spice Kitchen is building a content cluster about the history of the spice trade, not as a shop page but as a research resource that happens to lead people toward the brand along the way. Nathan Pearson from Lumos Digital said the focus is shifting from point-of-purchase to the research stage. That tracks with what we see in practice. Our testing across content clusters versus standalone product pages shows citation rates three to five times higher when genuine topical depth sits behind the content.

A UK solar installer we worked with had a single product page for battery storage that ranked on page two and got zero AI citations. We built a content cluster around it covering electricity tariff comparisons, export rate changes, G99 application timelines, and peak shaving calculations for specific DNO regions. Within four months, ChatGPT was citing the tariff comparison page in three out of ten test prompts for “is solar battery storage worth it UK.” The product page itself never got cited. The research content around it did.

E-E-A-T signals are the entry ticket. Andy Lochtie from Lumos Digital mentioned expertise, authority, and trust indicators, and he’s right that they’re no longer optional for any site that wants AI visibility. Inbound links from trusted sites, outbound links to quality sources, author biographies, and published content policies all feed into whether an AI system considers you citable. E-E-A-T authority signals appear in 96% of AI Overview citations according to Stacker’s December 2025 research.

What did the BBC miss?

The article gave a decent overview for a general audience but left out the mechanics that determine whether content gets cited or ignored.

Here’s where it gets interesting. And honestly, this is the stuff that separates sites getting AI citations from sites wondering why they’re invisible.

Does front-loading content really affect AI citation rates?

Yes. Analysis of 18,012 verified citations shows 44.2% reference the first 30% of page content. Only 24.7% come from the final third.

That’s from Kevin Indig’s Growth Memo study of 3 million ChatGPT responses (February 2026). The first third of your page does nearly half the work. The final third barely registers for citation purposes.

What does that mean for you? Put the answer right after each heading. The definition. The key number. The direct response. Don’t warm up. Don’t build suspense. Answer first. Elaborate after.

We restructured the top 15 pages on a client’s site to front-load answers in Q1 2025. Within three months, ChatGPT citation frequency for their target queries went from zero to appearing in four out of ten test prompts. The only change was content order.

Where does ChatGPT actually get its information?

ChatGPT’s source selection is more complex than most guides suggest, and it shifted dramatically in 2025 toward Google’s index and away from Bing.

This catches most people off guard. A year ago, the standard advice was to prioritise Bing because ChatGPT relied on Bing’s search infrastructure. That advice is now outdated.

Profound’s analysis of 240 million citations tracked ChatGPT’s index alignment between April and July 2025 and found Google alignment jumped from 12% to 33% while Bing alignment dropped from 26% to just 8%. Google Search Console verification and sitemap submission are now the primary levers for ChatGPT visibility. Still submit to Bing, but the priority has flipped.

And here’s what makes it even more complicated. ChatGPT’s premium model bypasses both search engines entirely. More on that in the next section, because it’s the single most important finding the BBC missed.

One thing that hasn’t changed: Bing Webmaster Tools launched an AI Performance dashboard in February 2026 that shows citation counts and grounding query phrases from Microsoft Copilot. It’s still the closest thing to a direct window into how AI selects sources. Set it up for every site you manage.

Why do premium and free ChatGPT users see completely different answers?

Free and premium ChatGPT users get fundamentally different business recommendations because the two models cite completely different sources with only 7% overlap.

This is the biggest finding the BBC missed, and it changes how every business should think about AI visibility.

Writesonic analysed 1,161 citations across 50 prompts in March 2026 and found ChatGPT’s two current models operate as essentially different citation universes.

GPT-5.3, the free model that most people use, sends just 8% of its citations to brand websites. It finds businesses through search engine rankings and primarily cites editorial and third-party sources. On comparison queries specifically, it cited zero brand websites. None.

GPT-5.4, the premium model, sends 56% of citations directly to brand websites. It breaks each question into an average of 8.5 sub-queries and uses site-specific search operators to query brand domains directly. No previous ChatGPT model did this. 75% of the domains GPT-5.4 cites don’t even appear in Google or Bing results for the same question.

Why does this matter? Premium ChatGPT users are the ones researching with intent to spend. They see a completely different answer to the same question than free-tier users. If your website publishes real pricing in plain HTML with clear service descriptions, GPT-5.4 will find it directly and cite it. If your site hides behind “call for a quote” or loads content through JavaScript frameworks, GPT-5.4 moves to a competitor who publishes that information.

The practical upshot: publish your prices. Ranges are fine. But “competitive pricing” means invisible to the buyers who matter most.

Do heading formats affect AI citation rates?

Content with question marks in headings generates double the citation likelihood, with 78% of cited heading content coming from question-format headings.

The BBC didn’t mention heading structure at all, which is surprising because it’s one of the easiest and fastest wins in AEO. Frame your H2s as the actual questions your audience asks, then answer immediately in the first sentence after the heading.

This connects to something called an answer capsule, which turned out to be one of the strongest citation predictors in recent research. Search Engine Land audited nearly 2 million sessions and found 72.4% of blog posts cited by ChatGPT included one. It’s a concise answer of roughly 20 to 25 words placed directly after a question-based heading, with no links inside it. That’s the format I’ve been using throughout this article, and the reason every bold sentence below a heading gives you the core answer before any elaboration.

Why wasn’t schema markup mentioned?

Only 12.4% of websites use structured data markup, and pages with JSON-LD plus FAQ blocks saw a 44% AI citation increase according to BrightEdge. The BBC didn’t mention schema once.

Schema markup tells AI crawlers what your content is about, who wrote it, when it was updated, and how it’s structured. Without Article schema, FAQ schema, and BreadcrumbList schema at minimum, AI has to work harder to understand your page. AI will always choose the path of least resistance. If your competitor’s page has clean schema and yours doesn’t, guess who gets cited.

And here’s a detail most guides skip entirely. AI crawlers don’t render JavaScript. They send an HTTP request, receive raw HTML, and move on. Erlin’s 2026 research measured parsing success rates across formats: static HTML with schema markup achieved 94% parsing success, JavaScript-rendered content managed just 23%, and PDFs scored 7%. If your schema is generated by client-side JavaScript, AI crawlers never see it. It needs to be in the server-rendered HTML.

For local businesses, the numbers are even more stark. Fewer than 5% of local business websites have any schema markup at all. That’s a massive competitive gap for anyone willing to do the technical work.

How much does content freshness affect AI citation rates?

Content updated within 30 days receives 3.2 times more AI citations than content older than 90 days, according to SE Ranking’s November 2025 research.

The BBC mentioned that MKM created new blogs. What it didn’t explain is that ongoing freshness keeps you in the citation rotation. About 45.5% of citations get replaced when the same query runs again according to position.digital tracking. Stale content loses its spot to someone publishing fresher answers.

But here’s the catch. Changing the date without changing substance only gives you a 12% citation bump. Adding new statistics to replace outdated ones gives you a 71% lift. AI knows the difference between a genuine update and a date change.

After 20 years in this industry, I’ll tell you that fake freshness has never worked long-term with Google and it works even less well with AI systems. Update with real substance or don’t touch it.

Which social platforms drive the most AI citations?

YouTube overtook Reddit as the top social citation source in late 2025, holding 39.2% of social citation share versus Reddit’s 20.3% according to Goodie AI’s analysis of 6.1 million citations from August to December 2025.

This one surprised everyone, and four independent research firms confirmed the shift in January 2026.

The counterintuitive part? OtterlyAI analysed over 100 million citation instances across six AI platforms in March 2026 and found 40.83% of AI-cited YouTube videos had fewer than 1,000 views. 36% had fewer than 15 likes. AI doesn’t care about audience size or engagement metrics the way social algorithms do. It cares about whether your video matches the topic and whether the content is structured clearly enough to extract a usable answer from.

On ChatGPT specifically, SEMrush analysed 325,000 prompts between January and February 2026 and found LinkedIn gets cited in 14.3% of responses. LinkedIn Pulse articles between 500 and 2,000 words perform best for citation.

If you’re only optimising your website and ignoring video and LinkedIn, you’re missing two of the three biggest citation channels outside traditional web content.

Why are the citation seats getting scarcer?

ChatGPT is citing fewer websites per response, dropping from an average of 19 unique domains to 15, which makes each citation slot harder to win and more valuable to hold.

That’s from Resoneo’s analysis of 27,000 ChatGPT responses tracked over 14 weeks through April 2026. The shrinkage happened after the GPT-5.3 model transition in early March.

To put that in context, AirOps analysed 548,534 pages across 15,000 prompts in March 2026 and found ChatGPT only cites 15% of the pages it retrieves. The other 85% get read and discarded. And only 30% of brands that appear in one response maintain visibility in the next.

This is a winner-takes-more dynamic. Fewer slots per answer means each one is fought over harder. SparkToro found in January 2026 that there’s less than a 1-in-100 chance ChatGPT gives the same brand list in any two responses. The recommendations rotate. But they rotate among a shrinking pool of sources that have earned enough trust across enough platforms to keep catching citations.

The businesses building that multi-platform presence now are the ones who’ll hold those slots as they get scarcer. Digital Bloom’s analysis of over 680 million citations found brands appearing on four or more third-party platforms are 2.8 times more likely to appear in ChatGPT responses. Erlin’s 2026 data shows brands with five or more active source types achieve 78% AI coverage. Brands relying on their own website alone manage just 18%.

What AEO mistakes should you avoid?

The biggest AEO mistakes are treating it as an SEO replacement, ignoring your multi-platform presence, and chasing AI traffic without fixing your technical foundation first.

After working with clients on this for over a year, these are the patterns we see killing AEO efforts before they start.

Dropping SEO for AEO. Google still handles the majority of search activity. The businesses winning right now do traditional SEO and AEO together, because AI citation feeds back into branded search volume and organic rankings over time. It’s a compound effect.

Having no presence outside your own website. 68% of AI citations come from third-party sources according to Erlin’s 2026 research. Only 32% come from brand-owned websites. If AI can only find information about you on your own domain, you’re invisible to most of the citation pipeline. Get profiles on Google Business Profile, Yelp, LinkedIn, Facebook, Bing Places, Apple Business Connect, and at least one industry directory. That gets you past the four-platform threshold where the 2.8x citation multiplier kicks in.

Writing for AI instead of writing for people. Weirdly, the content that performs best for AI citation is the content that reads most naturally to humans. AI answer engines are trained on human language patterns. Robotic, over-optimised content reads badly to both audiences. Graphite’s March 2026 data found 82% of content cited by ChatGPT and Perplexity is human-authored.

Skipping schema. Every client site we’ve audited that wasn’t getting AI citations had incomplete or missing schema markup. That includes sites with otherwise excellent content and strong organic rankings. It’s not glamorous work but it’s table stakes. And make sure the schema is in the raw HTML, not injected by JavaScript. At 23% parsing success for JS-rendered content versus 94% for static HTML, this isn’t a minor detail.

Publishing and forgetting. AEO rewards freshness more aggressively than traditional search ever did. Publish a piece and leave it untouched for six months and you’ll watch your citation frequency drop toward zero while competitors who update monthly take your spot in the rotation.

Hiding your pricing. This matters more than most businesses realise. ChatGPT’s premium model, the one used by people ready to buy, searches your website directly for pricing and service details. If it finds “call for a quote” instead of actual numbers, it moves to a competitor who publishes that information. Ranges are fine. Silence is not.

FAQ

Does AEO replace SEO?

No. AEO works alongside SEO and both are needed for full visibility in 2026. Google still handles the majority of search. AI citation compounds over time into branded search volume and organic rankings. The businesses winning right now build both systems in parallel. Expect a 6 to 12 month compound curve as citation visibility, brand mentions, assisted conversions, and direct traffic build sequentially.

How long does AEO take to show results?

Expect visible AI citation gains within 60 to 90 days, with meaningful traffic impact building over 6 to 12 months. You’ll see improvements in AI mentions first, then brand awareness, then conversion assists, and finally direct traffic. Clients who started 12 months ago now see AI delivering 7% to 15% of total traffic, and those visitors convert at more than double the rate of organic because they arrive pre-qualified.

Which AI platform should I optimise for first?

Start with Google AI Overviews because they appear in the search engine your customers already use. Then focus on ChatGPT. MKM Building Supplies confirmed in the BBC article that ChatGPT sends them more visitors than Google’s own AI features. ChatGPT also accounts for 77.97% of all AI referral traffic according to Conductor’s tracking.

Do small businesses need AEO?

Small businesses arguably need AEO more than large companies because the barrier to entry is low and conversion rates are high. The Spice Kitchen example from the BBC is more relevant than HubSpot for most businesses. A small company building topical authority to get discovered by AI searchers. You need good content, proper structure, schema markup, and patience. Not a massive budget.

What is the single most important AEO action I can take today?

Add answer capsules to your existing content by writing a 20 to 25 word direct answer after every question-based heading, with no links inside it. Research from Search Engine Land across nearly 2 million sessions shows 72.4% of ChatGPT-cited blog posts include this format. It takes an afternoon to retrofit your top pages and it’s the highest-impact change available for the least effort.

Can I automate AEO with a tool?

No. Tools help with research and tracking but AEO needs genuine expertise and first-hand experience signals that cannot be automated. If someone sells you an “AEO tool” that does it all, they’re selling the 2026 version of automated link-building software. The sites getting cited are the ones with real human expertise built into the content. That hasn’t changed and it won’t.

Where to start

If this feels like a lot, that’s because it is. But you don’t need to tackle everything at once, and the businesses getting results from AEO started with just a few focused changes before expanding.

Start with three things this week. Audit your top 10 pages and front-load the answers so each one gives the direct response within the first 150 words. Submit your sitemap to both Google Search Console and Bing Webmaster Tools. And add FAQ sections with proper JSON-LD schema to your five most important service or product pages, making sure the schema is in the server-rendered HTML rather than injected by JavaScript.

That’s a week’s work. Those three changes cover the highest-impact AEO actions available right now based on everything the citation research tells us.

Then build your presence across at least four third-party platforms. Google Business Profile, Yelp, LinkedIn, and one industry-relevant directory gets you past the threshold where citation probability nearly triples.

Everything else, the YouTube content, the LinkedIn articles, the monthly freshness updates, the cross-platform strategy, that’s the 6 to 12 month build that compounds over time. Start with the foundations and layer the rest in as you go.

If you want us to audit your site’s AEO readiness and show you exactly where citations are being left on the table, get in touch for a free analysis.

Source reference:https://www.bbc.co.uk/news/articles/c70n2rjgxeyo
https://www.kevin-indig.com/growth-memo/
https://www.brightedge.com/resources/weekly-ai-search-insights/ai-overviews-one-year-presence-size-citing