Introduction: The Search Box Isn’t the Only Front Door Anymore

For twenty-five years, the brief for search visibility was simple, if not easy: rank on page one of Google, and the traffic follows. That brief has quietly split in two.

A growing share of the questions your prospects used to type into Google are now typed, or spoken, into ChatGPT, Claude, Perplexity, and Google’s own AI Overviews. Your prospect isn’t looking at ten blue links and deciding which one to click. They’re reading a synthesised answer, generated in real time, that may or may not mention your brand at all.

The scale of that shift was one of the standout findings from BrightonSEO 2026, the industry’s leading search conference, which myself and our Director, Paul Thornton attended together in May. Jack Lingard opened one of the headline sessions with a genuinely sobering figure: 93% of AI-driven searches end without a click. Not a fifth. Not half. The overwhelming majority. And yet the same conference produced a statistic that reframes that number entirely rather than making it a reason to panic – data shared in Paul’s session with Nick Beck showed that visitors who do arrive at a website from AI search convert at roughly 4.4 times the rate of traditional organic visitors. Smaller volume, sharper intent.

That’s the shift this guide is about. We call it Generative Engine Optimisation, or GEO: the discipline of earning visibility and citations inside AI-generated answers, rather than simply ranking a webpage next to them.

This isn’t a replacement for SEO. It’s an extension of it, built on the same foundations – technical health, authoritative content, real-world reputation – pointed at a new set of gatekeepers. This guide extends two posts from our BrightonSEO 2026 recap series: our hub post on the future of search, and our deep dive on getting your brand mentioned by ChatGPT, Claude and Google’s AI. If you’ve read those, you’ll recognise the throughlines. If you haven’t, everything you need is below – we’ve pulled the most commercially relevant findings through into this guide directly.

By the end of this guide, you’ll understand:

What GEO actually is, and how it’s genuinely different from traditional SEO, not just a rebrand of the same tactics
How AI search engines retrieve, weigh, and cite information (to the extent this is publicly verifiable)
A repeatable content framework built for both human readers and AI summarisation
Platform-specific tactics for Google AI Overviews, ChatGPT, Perplexity, and Claude
How to audit your own brand’s current AI visibility, today, for free
Why digital PR, not link building, is the fastest lever you have for AI trust signals
A realistic 12-month action plan, with links to every supporting guide in this series

Book a free AI Visibility Audit with Digital Hothouse to see exactly how your brand currently appears (or doesn’t) across the major AI platforms – details at the end of this guide.

1. What GEO Is, and How It Differs From Traditional SEO

Generative Engine Optimisation is the practice of structuring your brand’s content, data, and online reputation so that generative AI systems can find it, trust it, and cite it when answering a relevant question.

The goal isn’t a ranking position. There is no page one. The goal is inclusion; being one of the sources an AI model draws on, and ideally being named, when it answers a question your buyer is asking.

Where GEO overlaps with SEO

Most of what makes a page rankable also makes it citable: clear structure, genuine expertise, fast and crawlable pages, credible signals of authority. If your technical SEO and content strategy are already sound (see Pillars 2 and 3 of this series), you are not starting from zero.

Where GEO genuinely diverges

Traditional SEO	GEO
Optimises for a ranking algorithm and a results page	Optimises for inclusion in a generated, synthesised answer
Success = clicks to your site	Success = being mentioned, cited, or quoted – a click is a bonus, not the goal
One primary ranking factor set (Google’s)	Multiple distinct systems (Google AI Overviews, ChatGPT, Perplexity, Claude, Gemini), each with different retrieval methods
Keyword and intent matching	Entity and fact matching – the model is trying to extract a verifiable claim, not match a phrase
Backlinks are a core trust signal	Brand mentions – linked or not – appear to carry weight, alongside citations and structured data
You control the destination page	You often don’t control whether you’re named at all – the model decides

The uncomfortable part of that last row is worth sitting with. In classic SEO, you could audit exactly why a competitor outranked you. With GEO, the reasoning inside a large language model’s answer generation is not fully transparent, even to the platforms themselves in some cases. This guide focuses on the levers that are observable and controllable, not on reverse-engineering a black box.

The number that should change how you prioritise

Christian Desert 56 Per Cent of AI Visibility from External Sources

If there’s one statistic from BrightonSEO 2026 worth building your GEO strategy around, it’s this one, from Christian Desert’s session on GEO: 56% of AI visibility influence comes from external sources. Your own website, however technically perfect, however well-written, accounts for less than half of what determines whether an AI platform mentions you at all.

This single number is why GEO can’t be run as a content-team side project bolted onto the CMS. It has to be a cross-functional effort spanning your website, your PR and comms function, and your reputation management, because the model doesn’t distinguish between those departments. It just reads the whole web and forms an opinion.

Jack Lingard’s companion finding at the same conference sharpens the point further: being cited in AI platforms is 6.5 times more likely if you’re already cited in third-party sources. Third-party citations aren’t a nice-to-have adjacent to GEO. For most brands, they’re the primary lever, which is why Section 6 of this guide, on digital PR, deserves as much of your attention as the technical sections.

Why this matters for revenue, not just visibility

If AI Overviews and chat-based answers are increasingly satisfying a query without a click, then two things are true at once: overall organic traffic to informational pages is likely to soften over time, and brand awareness delivered through an AI-generated answer still has commercial value, arguably more, since it arrives with implicit third-party endorsement from the platform. Pillar 5 of this series covers how to measure that value once clicks stop being the whole picture. For now, the point is this: absence from AI answers is not neutral. It’s a competitive gap.

2. How AI Search Engines Actually Retrieve and Cite Sources

This is the section where confidence should be proportional to what’s actually publicly documented. Christian Desert’s BrightonSEO session on GEO gave us the cleanest framework we’ve come across for explaining this to clients: there are fundamentally two distinct pathways by which your brand ends up in an AI answer, and they require different strategies.

Training data. The model already “knows” who you are because it was trained on content that mentioned you – your website, but also press coverage, reviews, forum threads, YouTube transcripts, and everything else that existed about you at the point the model was trained. This is a slow-moving, cumulative signal. You can’t influence it this week; you influence it over the coming months and years by building consistent, credible presence everywhere the model is likely to have learned from.

Retrieval-Augmented Generation (RAG). The AI pulls live content from the web at the moment of the query, layering that fresh information on top of (or instead of) what it learned in training. This is why you’ll often see citations or “sources” links attached to AI Overviews, Perplexity answers, and browsing-enabled ChatGPT and Claude responses.

Christian Desert How AI Answers are Reached

Paul’s take from BrightonSEO: “The RAG distinction was one of the most practically useful things I heard across the whole conference. If you can identify which of your queries are triggering live web lookups – typically time-sensitive, local, or fast-moving topics – you can prioritise those in your content strategy, because real-time content moves the needle far faster than waiting for the next training cycle.”

This has two direct implications for GEO strategy:

Being crawlable and indexable is not optional. If a retrieval-based system can’t access your page, it cannot cite you, no matter how good the content is. Janaina Barreto-Romero’s AI crawl analysis work, presented at BrightonSEO, found that websites relying heavily on client-side JavaScript rendering are frequently invisible to AI crawlers altogether. Unlike Google’s crawler, which has spent over a decade getting better at rendering JavaScript, many AI crawlers behave closer to Google circa 2012: they fetch the raw HTML and move on. If your key content only appears after JavaScript executes in a browser, it may simply not exist from an AI’s point of view. Pillar 2 of this series covers the full technical checklist.
Being in the training data still matters for baseline recognition. This is largely outside your direct control week-to-week, but it’s built cumulatively through sustained public presence: press coverage, Wikipedia and Wikidata entries, review platforms, and consistent brand data across the web. We cover this in the entity-building section below.

A verification note on llms.txt: you may have seen this proposed as an AI-optimisation quick win – a file that tells AI crawlers how to treat your content, similar in spirit to robots.txt. Thomas Peham from Otterly.ai ran controlled tests on this at BrightonSEO and found it isn’t worth your time: only around 0.1% of AI bot traffic accessed llms.txt files in his testing, with no measurable positive correlation to visibility. If someone recommends this as your AI optimisation strategy, that budget is better spent on the levers covered in Sections 3 and 6.

3. The GEO Content Framework: Answer-First Structure, Schema, Entities, and E-E-A-T

This is the operational core of GEO. Four pillars, applied together:

3.1 Answer-first structure

AI systems extract passages, not whole pages. Content that buries its answer under three paragraphs of preamble is harder to extract cleanly than content that states the answer immediately, then supports it.

Practically, this means:

Lead each section with a direct, self-contained answer to the question implied by the heading – a sentence that would make sense quoted in isolation
Use question-based H2/H3 headers that mirror how people actually ask (including how they’d phrase it to a chatbot, which is often more conversational than a typed search query)
Keep supporting explanation and nuance after the direct answer, not before it
Avoid front-loading brand narrative or scene-setting in sections meant to answer a specific question

This also happens to be better writing for human readers with limited attention. There’s no real tension between optimising for AI extraction and optimising for people – a topic we cover in depth in the Content Strategy pillar.

3.2 Schema and structured data

Structured data (Schema.org markup, implemented as JSON-LD) doesn’t guarantee an AI citation, but it removes ambiguity. It tells any system – search engine or language model – exactly what an entity, a fact, or a claim on your page actually is, rather than leaving it to infer from unstructured prose.

At minimum, a brand serious about GEO should have:

Organization schema on the homepage, with accurate sameAs links to verified social and knowledge-base profiles
FAQ schema on pages that genuinely answer discrete questions (don’t force it onto pages that don’t)
Author/Person schema with credentials, to support E-E-A-T signals (see below)
Review and Product schema where applicable to commercial pages

We go deep on implementation in the Technical SEO pillar’s schema markup guide – worth reading alongside this section.

Darko Brzica Entity Stacking

3.3 Entities over keywords

Traditional SEO thinks in keywords. GEO increasingly requires thinking in entities – the specific people, organisations, products, and concepts a system can look up, verify, and connect to other data.

This is why brand entity-building (Wikipedia notability, a maintained Wikidata entry, a verified Google Knowledge Panel, consistent citations across authoritative directories) has become a GEO tactic, not just a PR nice-to-have. A well-established entity gives an AI system something concrete to anchor a citation to. We cover this build process in a dedicated cluster post.

Darko Brzica’s BrightonSEO session on entity architecture introduced a concept we’ve adopted internally and now build into every client onboarding: the entity confidence score – an informal but useful way of thinking about how certain an AI system is about who you are. It’s built from consistency: the same brand name, the same description, the same core facts, appearing across Wikipedia, LinkedIn, your Google Business Profile, Wikidata, and the directories relevant to your sector. Schema markup that maps these together (technically, using sameAs properties to link your entity across platforms) is unglamorous work, but it’s foundational rather than optional.

Sean Barber, SEO Manager at Macmillan Cancer Support, made a point at BrightonSEO that has stuck with our team since: AI infers credibility from patterns across the web, not necessarily from truth. It looks for consistency – do the same facts and signals show up in multiple credible places? – and rewards brands whose narrative is coherent and widespread. It’s a slightly uncomfortable idea, because it means an inconsistent but honest brand can score worse than a consistent one, but it’s a useful discipline to build toward regardless: say the same true things about yourself, everywhere, all the time.

3.4 E-E-A-T, applied literally

Google’s Experience, Expertise, Authoritativeness, and Trust framework was written for human quality raters, but it maps closely onto what AI systems appear to weight when selecting sources to cite. In practice:

Experience: first-hand case studies, original data, and specific outcomes – not generic advice that could have been written by anyone
Expertise: named authors with visible, verifiable credentials
Authoritativeness: third-party validation – press mentions, citations from other credible sources, industry recognition
Trust: accuracy, transparency (clear authorship, dates, sourcing), and a track record of not needing correction

None of this is new advice. What’s changed is the audience: you’re no longer just writing to satisfy a quality rater guideline. You’re writing to be the source a model reaches for when it needs to sound credible.

4. Platform-by-Platform Tactics: Google AI Overviews, ChatGPT, Perplexity, and Claude

Each platform behaves differently enough that a single generic strategy underperforms. Broad directional guidance below – verify current mechanics for each platform before building a campaign around them, as these systems are updated frequently.

Google AI Overviews

Draws heavily on Google’s existing index and ranking signals, which means traditional SEO fundamentals (crawlability, structured data, strong topical relevance, backlink profile) directly influence inclusion likelihood. Content that already performs well for featured snippets tends to be a strong candidate. Schema markup and clear, concise answer blocks appear to help.

ChatGPT

Behaviour depends heavily on whether a given response is using browsing/retrieval or drawing from training data. For browsing-enabled responses, standard crawlability and content clarity apply. For training-data-based recognition, sustained public presence over time (press, reviews, structured entity data) matters more than any single piece of content you publish this month – this is a long-game lever, not a quick win.

Perplexity

Built around live retrieval and citation by design, it’s the platform most explicitly built to show its sources. This makes it a good testing ground: search your brand’s target queries directly in Perplexity and you can see, in real time, which sources it’s pulling from and whether you’re among them.

Claude

Similarly capable of live retrieval when connected to search, with a comparable emphasis on grounding answers in identifiable sources. As with ChatGPT, baseline brand recognition without browsing draws on training data and prior public presence. One specific, actionable finding from Sam Davis’s BrightonSEO session on LLM brand management: Claude appears to place particular weight on review response cadence as a trust signal for local and product queries, not just whether you have reviews, but whether you’re actively and promptly responding to them. A brand whose last review response was eight months ago is sending a signal, whether it means to or not.

The common thread

Regardless of platform-specific quirks, the same underlying assets keep coming up: crawlable and well-structured pages, clear entity data, genuine third-party validation, and content that states things plainly enough to be extracted and trusted. It’s also worth knowing where these models are learning from in the first place: Thomas Peham’s data at BrightonSEO found YouTube to be the second most cited social platform in AI responses, after Reddit – both worth a deliberate presence if your category lends itself to video or community discussion. Chase the fundamentals rather than chasing any single platform’s current algorithm – the fundamentals are far less likely to be obsolete in six months.

5. How to Audit Your Current AI Visibility

You can start this today, for free, in about thirty minutes. This is the same mirror-test methodology we run for clients, refined from the process we brought back from BrightonSEO. It’s a lightweight version of the full framework in our dedicated AI Visibility Audit cluster post – useful as a first gut-check.

Step 1: The basic mirror test. Open ChatGPT, Claude, Gemini, and Perplexity. Ask each one: “What is [your brand]?”, “What does [your brand] offer?”, “Is [your brand] a good [type of business]?”, and “What do customers say about [your brand]?” Record whether each answer is accurate, current, and whether it names the right products, services, and location.

Step 2: Category queries, brand name removed. Ask “What are the best [your category] in [your location]?” and “Who are the leading [your type of business] for [your target customer]?” without naming yourself. Do you appear? In what context? If you don’t, which brands do – and what do they visibly have that you don’t?

Step 3: Sentiment and source check. Enable web browsing in ChatGPT and Perplexity where available, and re-run the category questions with browsing on. This is the RAG layer in action, and the results can genuinely differ from the training-data layer. Note which sources get cited – those are the publications and platforms that matter most in your category, and a strong signal for where to focus digital PR effort.

Step 4: Review your external presence. Check Google Reviews, Trustpilot, your most relevant sector directories, and do a quick Reddit search for your brand name. Are there negative or outdated threads that have never been addressed?

Step 5: Entity consistency check. Search your brand across Google Business Profile, LinkedIn, Wikipedia or Wikidata, and any major sector directories. Is the name spelt and formatted consistently? Are descriptions and locations accurate everywhere? Inconsistency here is what suppresses the entity confidence score covered in Section 3.

*Bonus Tip: when running these AI Mirror tests, turn on Incognito mode within each of the AI platforms. That way, your results are not influenced by your previous searches, creating a more neutral response that is likely to me more in line with what others see (although their results will also be influenced by their own search history). The Incognito button is usually found in the top right across all platforms.

On tools: a few platforms came up repeatedly at BrightonSEO for ongoing AI-visibility tracking – SE Ranking for LLM visibility and sentiment tracking, Waikay for prompt-level tracking, Oncrawl for log-file analysis of which AI crawlers are actually visiting your site, and Kompote.ai for AI inclusion tracking. We’re still evaluating several of these ourselves, and the “right” toolset a year from now will likely look different again.

Paul’s take: “None of these tools are perfect yet. The more important shift is in mindset – regularly checking how AI platforms represent your brand should become as routine as checking your rankings or your review score. You don’t need a tool to start doing that. You just need to make it a habit.”

If you’d rather this was done properly, with a structured methodology and a written report you can take to your board, book a free AI Visibility Audit with Digital Hothouse.

6. GEO + Digital PR: Why Brand Mentions Matter More Than Backlinks

This is arguably the most important strategic shift in this whole guide, and it’s the one most marketing teams are slowest to act on.

Traditional link building optimised for one thing: a hyperlink, ideally from a high-authority domain, pointing at your site. That’s still valuable. But AI systems appear to weight something broader: the frequency, consistency, and credibility of your brand being mentioned across the web, whether or not those mentions are hyperlinked. As covered in Section 2, that 56% external-influence figure is the reason this matters as much as it does.

Thomas Pelham Power of Digital PR

James Roach put it plainly at the Digital PR Summit, in a line worth repeating to any board that still treats PR and SEO as separate budget lines: digital PR in 2026 is not a separate channel from SEO. It is SEO conducted outside your own website. Every piece of press coverage that mentions your brand is a third-party citation. Every editorial roundup that includes your product is an entity signal. Every journalist who quotes your team as an expert source is an E-E-A-T signal.

What makes this actionable rather than just directionally true is a specific finding from Thomas Peham’s controlled testing at Otterly.ai: distributing press releases to trusted media outlets produced measurable AI citation uplift within days of publication, not weeks, not months. For a channel that usually gets valued on slow-burn brand awareness, that’s a surprisingly fast feedback loop.

How to actually run this, based on what worked at BrightonSEO:

Go narrow, not wide. Vince Nero’s research at the Digital PR Summit found that hyper-targeted campaigns of 25–50 highly personalised pitches generated 17 times more coverage than mass outreach to thousands of generic contacts. Build a shortlist of journalists who’ve genuinely covered your category recently, and pitch them properly rather than blasting a database.
Lead with original data. Multiple speakers converged on the same point: the only truly irreplaceable PR asset in an AI-saturated content landscape is exclusive data nobody else has. George Sinnott’s framework is the one we now use with clients – start with a question people suspect the answer to but that nobody has actually measured, then find the data. The strongest stories are relatable, visual, quantified, surprising, and tied to real consumer impact.
If you sell physical products, chase editorial roundups specifically. Amy Gibson’s session at the Digital PR Summit showed that “10 best gifts under $100”-style roundup articles are now among the most heavily cited content formats when someone asks an AI for a product recommendation. Securing placement in relevant roundups is no longer just a coverage win, it’s a direct AI-visibility play for retail and e-commerce brands.
Treat reviews as a PR asset, not just a support function. Old, unresolved reviews and negative Reddit threads don’t age out – Janaina Barreto-Romero’s research showed AI systems are still citing years-old reviews today. Combined with Claude’s specific weighting of review-response cadence (Section 4), an active, well-managed review programme is now doing double duty as a reputation asset and a GEO signal.

Practically, that means the campaigns covered in our Digital PR pillar- original data studies, expert commentary, newsjacking, journalist relationships- aren’t just brand-awareness or backlink plays anymore. They’re direct inputs into your AI visibility. If you’re running SEO and PR as separate workstreams with separate KPIs, this is the argument for bringing them under one strategy.

Paul’s take, from the Digital PR Summit: “Michael Bates’s session on brand narrative was one of the few that made me genuinely uncomfortable about our own work, because I could think of clients immediately where a proper audit would reveal real inconsistencies between what their website says and what’s actually being said about them elsewhere. The good news is that audit is quick to run, and it gives you a clear, prioritised action list.”

7. Your 12-Month GEO Action Plan

You don’t need to do all of this at once. Here’s a realistic sequence, with links to the supporting guides in this series as they publish.

Months 1–2: Foundation and audit

Run the AI visibility audit above and establish your baseline
Confirm technical crawlability for AI bots – see the Technical SEO pillar
Audit existing schema markup and close obvious gaps

Months 3–5: Content structure

Rework your highest-priority pages into answer-first structure
Roll out FAQ, Organization, and Author schema across priority pages
Begin building or reinforcing entity signals (Wikidata, Knowledge Panel, directory consistency)

Months 6–8: Authority building

Launch or scale a digital PR programme aimed at earned, credible mentions
Strengthen author bios and credential visibility for E-E-A-T
Re-run the AI visibility audit and compare to baseline

Months 9–12: Measurement and iteration

Establish ongoing monthly AI-mention tracking (see the Measurement pillar for dashboard approaches)
Double down on the platforms and query types where you’re gaining traction
Fold AI visibility metrics into standard marketing reporting, not a separate side project

Supporting guides in this pillar

What Is Generative Engine Optimisation (GEO)? A Beginner’s Guide for NZ Businesses
How Google AI Overviews Work (And How to Get Featured In Them)
ChatGPT vs Google: How Different AI Platforms Choose Which Brands to Mention
The Content Structure That Gets Cited by AI: Answer-First Writing Explained
Schema Markup for AI Search: What You Need in 2027
Building Brand Entities: Why Wikipedia, Wikidata and Knowledge Panels Matter More Than Ever
How to Audit Your Brand’s Current AI Visibility (Free Framework)
E-E-A-T in the AI Era: Proving Experience, Expertise, Authority and Trust to Machines
Local Business AI Visibility: Getting Mentioned in AI Search for “Near Me” Queries

Get Your Baseline Before You Build a Strategy

Everything in this guide compounds. Schema helps discovery, which supports citation, which is reinforced by PR-driven mentions, which builds the entity signals that make training-data recognition more likely, but none of it is worth prioritising blind. You need to know where you actually stand today.

Book a free AI Visibility Audit with Digital Hothouse. We’ll test your brand against real buyer queries across ChatGPT, Claude, Perplexity, and Google AI Overviews, benchmark you against competitors, and hand you a prioritised action list, not just a report.

News and insight

category: AI Optimisation

How Do You Build Trust and Credibility for AI-Powered Search?

News and
insight