Edge Studio · AEO Framework v2.0
The core insight
Search engines rank everything and let humans choose. Agent systems do something different — they decide first, before any recommendation, which brands they trust enough to mention at all. That decision is the Eligibility Layer.
Most brands assume AI knows who they are. It does, in the loosest sense. What it does not have is enough structured, consistent, credible information to confidently recommend them over competitors.
That gap is the problem AEO solves. Not ranking. Not content volume. The binary question of whether your brand is inside the candidate set the model will consider at all — and then the gradient question of how strongly it reaches for you when it does.
You cannot rank inside a system that has already decided not to mention you.
The framework
01 / First condition
Binary · Fixable
The model can accurately describe your brand — what it is, what it does, who it serves — without ambiguity or contradiction. This is an entity problem. AI builds internal representations of brands. A brand that has not established a clear, consistent entity definition is, to the model, ambiguous. Ambiguous entities are not recommended.
Without it: AI ignores the brand entirely or generates inaccurate descriptions that reduce recommendation confidence.
02 / Second condition
Threshold + gradient · Buildable
The model has encountered your brand in enough credible, relevant contexts to include it in the candidate set for recommendation. Trust is not sentiment — it is a structural property of how often and where the model has seen your brand mentioned. Citation frequency, factual consistency, co-occurrence with sources the model already weights as credible.
Without it: The brand is understood but not surfaced. It exists in the model's knowledge but not its recommendations.
03 / Third condition
Gradient · Competitive · Ongoing
When two trusted brands compete for the same recommendation, the model reaches for yours first. This is where most AEO frameworks stop short — and where the majority of actual recommendation outcomes are determined. Preference is built by accumulating a lead: share of voice in category discussions, problem-solution co-occurrence, authority amplification at scale.
Without it: The brand is eligible but loses the selection to a competitor with stronger category authority signals.
The three conditions are sequential. A brand that is not Understood cannot be Trusted. A brand that is not Trusted cannot be Preferred. But passing the first two does not guarantee the third. That is where most brands stall.
The mental model
There is a natural sequence: you cannot earn AI trust before you have given AI something clear to understand. You cannot build preference before you have built trust. The conditions are sequential and they matter in order.
But the accurate mental model is a flywheel. Each Answer Hub update sharpens entity clarity. Each new third-party citation raises the trust threshold. Each piece of category-defining content strengthens preference signals. The compounding starts when the foundation is in place and you keep turning the wheel.
The brands that dominate AI recommendations are not the ones who built it once. They are the ones who kept compounding.
AI builds internal representations of entities — companies, products, categories. A brand that has not established a clear, consistent definition across the surfaces AI learns from is, to the model, ambiguous.
The test: ask ChatGPT, Perplexity, and Gemini to describe your brand without prompting. Every discrepancy between the response and your actual positioning is a signal failure — and a specific fixable problem.
This layer is binary. Either the model has a clear picture or it does not. The Foundation work fixes it directly.
AI models build mental models of entities. If your brand does not exist as a recognisable entity in the web's knowledge graph, AI has low confidence recommending you regardless of how good your content is. Fix this before you build anything else.
Build a living Answer Intent Map across four AI models: ChatGPT, Perplexity, Gemini, and Claude. A brand can be dominant in Perplexity and invisible in ChatGPT. You need parity across all four, or at minimum, you need to know where the gaps are.
Every significant page on your site should be built to answer a specific question a human would ask an AI. URL structure, H1s, first-100-word TL;DRs, FAQ sections, and an llms.txt file telling AI crawlers what to prioritise.
The central content asset for AI citations. Include a "Last reviewed" date and update it regularly. AI models weight content freshness. A guide updated last week will outcompete a guide updated 18 months ago even with similar content.
A machine-readable source of truth for your brand. Every fact should be identical everywhere: your website, Amazon, Google Merchant Center, social bios. Cross-reference your Brand-Facts page with your Answer Hub to build the internal trust loop AI crawlers reward.
Standard schema plus SpeakableSchema on TL;DR paragraphs and FAQ sections. SpeakableSchema explicitly signals to AI that this content is designed to be spoken or quoted, increasing the probability of direct citation.
Trust, in the context of AI recommendation systems, is not sentiment. It is a structural property of how often and in what contexts the model has encountered your brand, and whether those contexts were themselves credible.
A model learns trust signals the way a researcher evaluates sources: is this brand mentioned by people and publications I already trust? Is the information consistent across sources? This is Machine-Readable Reputation — and most brands have not started building it.
A threshold must be crossed to enter the candidate set. More trust signals continue to help beyond that threshold, but there is a minimum below which the brand simply will not surface.
AI models, especially Perplexity and ChatGPT, weight content from sources they associate with credibility. Earned media in category roundups builds more trust signal than generic backlinks. Review language matters too: AI reads the text of reviews and learns what customers say your brand does. Coach customers at the post-purchase stage to describe specifically what improved.
AI aggregates information from everywhere. If your product description on Amazon says one spec and your site says another, the model notices. Inconsistent facts actively reduce recommendation confidence — the model learned the inconsistency during training and suppresses your brand as a result. This is unglamorous work. It is also why one brand gets recommended and the identical brand next to it does not.
This is the condition most AEO frameworks do not address. Being Preferred means that when the model constructs a recommendation, your brand is the one it reaches for first — not because you are the only trusted option, but because the weight of signals associates you more strongly with the problem being solved than any competitor.
Preferred is not a state you achieve and maintain. It is a lead you build and defend. Competitors are publishing, earning coverage, and accumulating citations continuously. The brands that stay recommended are the ones that keep compounding.
Optimising specifically for AI shopping queries and purchase-intent prompts. This is where AI visibility converts to revenue. The brands recommended in "best X for Y" queries own the category in the moment that matters most — when a customer has decided to buy and is asking AI what to buy.
The most overlooked layer. Ask ChatGPT, Perplexity, and Gemini: "What is [Brand]? Is [Brand] trustworthy? What do people say about [Brand]?" This reveals the narrative AI has constructed about you. You might be eligible for recommendations but AI is describing a version of you that is 18 months out of date. Fix the narrative before you amplify it.
The Understood and Trusted layers are a structured project with a defined end state. The Preferred layer is an ongoing programme — sustained PR at scale, founder positioning as a named category expert, original research that earns citations, and authority amplification that compounds over time. This is the correct framing for an ongoing AEO partnership. Ask us about the Authority Programme.
AEO work increases the probability of AI recommendation. AI systems are non-deterministic and continuously updated — no methodology guarantees a specific outcome. What we can guarantee is that brands with stronger Understood, Trusted, and Preferred signals are recommended more often, more accurately, and more durably than brands without them.
The client journey
The Foundation work builds the structural layer. The Authority work builds the lead. Then the flywheel takes over.
Before you start
20 purchase-intent prompts. Score out of 20. You see exactly where your brand stands across Understood, Trusted, and Preferred — and which condition is the binding constraint.
Weeks 1 to 4
Layers 0 to 7. Entity clarity, structured content, schema, citations, consistency. The structural work that makes recommendation eligibility possible.
Month 2 onward
Layers 8 and 9 plus ongoing amplification. Category share of voice, founder positioning, PR at scale. This is where eligibility becomes preference.
Month 3 onward
Every citation, content update, and authority signal compounds. Monthly audit re-runs track score movement. This is where the ROI becomes unmistakable.
Common questions
The Eligibility Layer is the binary threshold AI systems apply before deciding which brands to recommend. Search engines rank everything and let humans choose. Agent systems decide first — which entities they trust enough to mention at all. That decision happens before any recommendation. Brands either pass the threshold or they don't. AEO is how you make sure you pass it.
These are the three conditions for passing the Eligibility Layer. Be Understood means AI can accurately describe your brand. Be Trusted means AI has encountered your brand in enough credible contexts to include it in the recommendation candidate set. Be Preferred means when two trusted brands compete for the same recommendation, AI reaches for yours. Each condition has different signals, different tactics, and a different relationship to the final outcome.
SEO targets page-one rankings in a list of ten blue links. AEO targets recommendation eligibility — the question of whether an AI system will mention your brand at all when someone asks for a recommendation in your category. The mechanics are fundamentally different: SEO rewards backlinks and keyword density. AEO rewards entity clarity, factual consistency, and Machine-Readable Reputation — the algorithmic credibility signals AI systems use to evaluate trust before deciding who to recommend.
An AI Visibility Score measures how many of a brand's highest purchase-intent queries result in an AI recommendation. Edge Studio's Found By AI tool runs 20 purchase-intent prompts across AI engines and scores brands out of 20. It is the core metric we track because it directly reflects the commercial impact of AEO work — and it is the baseline every engagement starts from.
The Foundation work — building the Understood and Trusted layers — is a 4-week structured project. Most brands see measurable AI visibility improvement within 60 days. The Preferred layer is ongoing: it is a lead you build and compound over time, not a state you achieve once. The compounding effect typically becomes significant from month three onward.
No. The two disciplines reinforce each other. Strong domain authority from SEO creates a foundation AI trusts. Strong AEO signals — entity clarity, structured data, consistent facts — improve how both search engines and AI systems understand your brand. The brands building both in parallel are the ones that will own their category in search and AI simultaneously.
Coming soon
We are building out the full implementation kit — templates, checklists, and everything you need to run the framework yourself. Join the waitlist and you will be first to know when it is ready.
Be first to get the full kit when it drops. One email. No sequences.
No spam. Unsubscribe any time.