Updated May 2026: Added 'What we've seen testing 75+ UK brands in 2026' section with four worked examples from real April 2026 SearchIntel runs (AJ Bell, Mercer alias, pension batch, Gemini methodology). Removed the unverified '30% improvement' schema-accuracy claim attributed to Canel and the unsourced 91/9 LinkSurge stat.
Search "AI SEO services" and you'll find two very different things bundled under the same label. Some agencies use AI tools to do traditional SEO faster, automating keyword research, content drafting and technical audits. Others focus on making your brand appear when someone asks ChatGPT, Gemini or Perplexity a question about your industry.
Both are valid. But they're not the same thing. And understanding where they overlap, and where they don't, is the difference between a complete strategy and a partial one.
The Good News: Strong SEO Foundations Transfer
If you've invested in solid SEO, you haven't wasted your time. Many of the fundamentals that help you rank on Google also help AI models find and trust your content.
The data confirms this. According to Ahrefs' research, 76% of Google AI Overview citations come from pages already in Google's top 10. A 2025 AI Visibility Report found that ChatGPT's citations show an 87% correlation with Bing's top 10 results. When these AI tools search the web in real time, they're pulling from the same index your SEO efforts are designed to influence.
Here's what carries over directly:
| SEO Foundation | AI Visibility Impact |
|---|---|
| Structured data & schema | Microsoft's Fabrice Canel confirmed at SMX Munich 2025 that schema markup helps Bing's LLMs and Copilot understand page content. Specific lift figures vary by site. |
| E-E-A-T signals | LLMs favour entities demonstrating expertise, authority and trust, the same signals Google rewards. |
| Content quality & depth | AI avoids thin, generic content. Original research and unique data increase AI visibility by 30-40%. |
| Crawlability | AI crawlers don't render JavaScript. Clean HTML, sitemaps and logical site architecture are prerequisites. |
| Content structure | Clear headings, lists and semantic HTML help LLMs extract and cite information accurately. |
| Internal linking | Forms your site's knowledge graph. Helps AI crawlers follow context and find supporting evidence. |
| Content freshness | Perplexity's Sonar model favours recently updated content. Even minor edits reset the freshness signal. |
As Search Engine Land put it: "The fundamentals haven't disappeared. What they've done now is matured. Strong technical foundations and quality content still underpin performance."
So if your SEO is strong, you have a genuine head start.
The Gap: Where SEO Alone Falls Short
Here's where it gets interesting. While strong SEO gets your content into the pool AI draws from, it doesn't guarantee you'll be cited.
A study by Chatoptic found that brands ranking on Google's first page were mentioned in ChatGPT only 62% of the time. And when the same brand appeared in both Google and ChatGPT, the rank correlation was nearly zero, just 0.034. Ranking #1 didn't mean being mentioned first. Or at all.
Search Atlas analysed authority metrics across LLM responses and found correlations between domain authority and AI visibility ranged from -0.08 to -0.21. Traditional authority signals had limited influence on how often domains appeared in AI answers.
Why? Because AI models aren't just pulling from search results. They're synthesising information from across the web and applying their own trust criteria. Being retrievable is necessary, but it's not sufficient.
What we've seen testing 75+ UK brands in 2026
The third-party data above tells the broad story. Here's what SearchIntel has measured directly across our own monitoring runs since February 2026, testing UK brands across pensions, comparison sites, travel, recruitment and publishing.
Same brand, same SEO, two prompt sets, 84-point swing
We tested AJ Bell, a FTSE 250 investment platform with strong SEO foundations and a top-3 organic position on its core terms, using two different prompt sets a week apart. The first set used generic wealth-management language ("best wealth management UK", "best UK financial adviser"). AJ Bell appeared in 8% of responses across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. The second set used investment-platform language ("best UK investment platform", "cheapest UK stocks and shares ISA", "best UK SIPP provider"). Same brand, same week, same SEO. Score: 92%.
SEO didn't change between runs. The query language did. AI models had clear category boundaries that traditional rank tracking would never surface. If you ask the wrong question, the brand looks invisible. Ask the right one, it's dominant.
Brand-name aliases matter more than they should
Across a 20-firm UK pensions batch, we initially saw Mercer UK score 0/100, bottom of the table, while "Mercer" was the most-cited brand inside the same model responses. The model knew Mercer. Our matcher didn't, until we registered the short form as an alias. The corrected score: 88/100. Same data, same brand, same Google rank.
Traditional SEO never had this problem. URLs and rankings are exact-match. AI's mental model of your brand is fuzzy and shorthand-driven. If you trade as "Compare the Market" but the AI says "CtM" or "Compare", measurement that doesn't allow for that returns false zeros. Naming variance is now an audit input.
Vertical signals beat firm-level SEO
The same 20-firm pensions batch ran identical keywords, same five platforms, same three-runs-per-prompt methodology. Scores ranged from 0/100 to 100/100. The top quartile (NEST 100, Mercer 88, Aviva 88, Hymans 88, Barnett Waddingham 88) shared one trait: heavy third-party citation density via Wikipedia, government source pages, regulatory bodies and trade press. The bottom quartile (Capita 0, Spence and Partners 4, Buck 8) had comparable Google authority but thinner off-site presence in AI-trusted sources.
Pages can rank. Brands need to be cited. Different game.
Methodology matters as much as content
We caught a tooling issue in late April 2026 where Gemini 2.5-flash was returning ~200-character truncated responses across an entire batch, the model was burning its output budget on hidden reasoning tokens before generating visible text. Disabling the thinking budget produced 5,500-character responses on the same prompts. Our visibility scores rose from artificially zero to accurate 80-100 across the affected firms.
The lesson is broader than one platform bug. If you're consuming AI visibility data without seeing the methodology, the prompts, the run count, the platform configurations, the alias rules, you're consuming a number that may or may not reflect reality. SEO has 25 years of standardised measurement. AI visibility doesn't. Vendor opacity is a risk worth pricing in.
How AI Platforms Actually Find and Choose Sources
Each AI platform has its own approach to sourcing information, but they all go beyond traditional search rankings:
- Google AI Overviews have the strongest tie to traditional search, 93.67% of citations link to at least one top-10 organic result. If you rank well on Google, you're well-placed here.
- ChatGPT searches via Bing when browsing is enabled. It shows 87% correlation with Bing's top 10, but only 11% of websites earn citations from both ChatGPT and Perplexity, each platform evaluates sources differently.
- Perplexity searches its own proprietary index of 200+ billion URLs. It shows about 60% overlap with Google's top 10, but prioritises recency and source diversity.
- Gemini and Claude rely more on training data plus real-time retrieval, weighting entity consistency and depth.
An analysis of 23,000+ AI citations by Omniscient Digital found that brand search volume was the strongest predictor of citations (0.334 correlation), followed by multi-platform presence across 4+ channels. Pages with schema markup were 3x more likely to earn AI citations.
The Additional Layer: What You Need Beyond SEO
This is where the strategies diverge. Traditional SEO is largely about your own website, content, technical health, backlinks. AI visibility adds a layer that's mostly about what happens off your website.
The pattern across credible studies is consistent. On branded queries specifically, analysis of 23,000+ AI citations by Omniscient Digital found the vast majority come from earned media, reviews, news articles, forum discussions and industry coverage rather than the brand's own website. Your owned content gets you in the pool. What gets you cited is what others say about you.
Here's what's new:
1. Third-party citations are now dominant
What others say about your brand matters more than what you say about yourself. Across major AI engines, Reddit consistently ranks as one of the single most-cited domains, around 40% of citations across ChatGPT, Google AIO, Perplexity, Gemini and Claude per the Semrush 150,000-citation analysis and the 5WPR AI Platform Citation Source Index 2026. Wikipedia consistently appears in the top three. The share is volatile: ChatGPT's Reddit citation rate dropped sharply in late 2025, then recovered. Reviews on Trustpilot, Google Maps and industry platforms also carry significant weight. None of this is something traditional SEO addresses.
2. Entity consistency across platforms
AI models cross-reference your brand across your website, Google Business Profile, LinkedIn, Crunchbase, industry directories and review sites. When messaging differs across platforms, AI interprets inconsistency as uncertainty and deprioritises the brand. Traditional SEO doesn't require this kind of cross-platform alignment.
3. Answer-first content structure
This builds on good SEO practice but takes it further. AI models extract direct answers. Our data shows Google AI Overviews open with a direct fact 99.8% of the time, zero preamble. Content needs to lead with the answer, not build up to it. Structure for extraction, not just for reading.
4. Cross-platform monitoring
Each AI platform behaves differently. Our research across our category research shows a brand can appear in 82% of Gemini responses and 0% of AI Overviews. You can't manage what you don't measure, and traditional rank tracking doesn't cover this.
The Click Landscape Is Changing Too
There's an urgency to this that goes beyond future-proofing. The value of traditional rankings is shifting.
A Seer Interactive study across 3,119 queries found that when Google AI Overviews appear, organic CTR drops 61% and paid CTR drops 68%. Ahrefs reported that AI Overview presence reduces clicks to the top-ranking page by 58%, and that figure is worsening over time.
Meanwhile, the AI search audience is massive and growing. ChatGPT has 900 million weekly active users. Perplexity processes over 60 million queries per day. Gemini has 750 million monthly active users.
But here's the critical point: brands being cited inside AI Overviews earn 35% more organic clicks than those that aren't. Being mentioned in the AI answer is now more valuable than ranking beneath it. Strong SEO puts you in contention. AI visibility strategy gets you cited.
A Framework: SEO as Foundation, AI Visibility as the Next Layer
Rather than thinking of this as "SEO vs AI visibility," think of it as two layers of the same strategy:
| Layer 1: SEO Foundations | Layer 2: AI Visibility | |
|---|---|---|
| Goal | Be findable and rankable | Be citable and trustworthy to AI |
| Focus | Your website | Your brand across the web |
| Tactics | Technical SEO, content, schema, E-E-A-T, backlinks | Third-party citations, entity consistency, review management, forum presence |
| Measurement | Rankings, traffic, conversions | AI citation rate, share of voice, recommendation rate |
| Who controls it | Mostly you | Mostly others (earned media, reviews, community) |
Layer 1 without Layer 2 means you rank on Google but get ignored by AI. Layer 2 without Layer 1 means AI can't find or trust your content in the first place. You need both.
What to Ask Your Agency
If you're evaluating agencies for "AI SEO services," here are the questions that reveal whether they're covering both layers:
- "How do you measure AI visibility?", If the answer is just Google rankings, they're on Layer 1 only.
- "Which AI platforms do you track?", ChatGPT, Gemini, Claude and AI Overviews all behave differently. A complete strategy monitors all of them.
- "What's your approach to third-party citations?", This is where Layer 2 lives. Reviews, Reddit, media mentions, industry coverage.
- "Can you show me our current AI citation rate?", If they can't measure it, they can't improve it.
The right answer isn't "we do SEO" or "we do AI." It's "we build the SEO foundations that make AI visibility possible, then add the strategies that get you actually cited."
the takeaway
SEO isn't dead. It's the foundation. But it's no longer the complete strategy.
The brands winning in 2026 are the ones treating SEO and AI visibility as complementary layers, using strong technical foundations, quality content and E-E-A-T to be retrievable, then building the third-party presence, entity consistency and citation-worthy content that makes AI trust and recommend them.
The question isn't whether to do SEO or AI visibility. It's whether you're doing both.