Okay, I need to talk about something that's been quietly bothering me for weeks. You know that feeling when you realise someone's been standing behind you reading over your shoulder, and you're not sure how long they've been there? That's basically what Google just did to AEO.

Here's the setup: for the past couple of years, everyone working in AEO has been operating from a fairly clean playbook. Make your content legible to AI systems. Structured data, topical authority, editorial credibility, engagement signals. We've covered it extensively in this newsletter, and the core argument has been consistent — AEO is fundamentally SEO with a new acronym. The underlying work didn't change. Just the three-letter abbreviation on the conference badge.

That argument still holds. Mostly. But Google just added a layer that the existing playbook doesn't account for, and it's significant enough that I've been staring at my ceiling about it. Because Google isn't just answering queries anymore. It's answering queries based on who's asking them.

At Google I/O last year, Google announced what they're calling context-aware search — a system that answers queries based not just on the words in the search box, but on the identity, history, and inferred intent of the person typing them. Right now, this personalisation layer lives inside Gemini and AI Mode. It hasn't migrated to AI Overviews yet — the surface most SEO professionals are still obsessing over. But the trajectory is about as subtle as a freight train. Personalisation is heading toward all of search.

Previous attempts at personalised search were, let's be honest, a bit rubbish. Google would look at your search history, sprinkle in some browsing data, and give you a slightly fancier version of "near me." Groundbreaking stuff.

What they've built this time is architecturally different. It runs on the same infrastructure as Google Alerts — the notifications system internally. Every search you run plants a flag telling the system you're interested in that topic. Those flags accumulate. Every future search gets answered in the context of every interest you've ever demonstrated, and that history becomes the background context for every AI-generated answer.

But here's where it gets properly creepy: it goes well beyond search history. Google pulls in calendar invites to understand who you meet with, combines that with Gmail signals, Maps history, and every behavioural trace you've left across Google's products over years of use. Android users get their call history and contacts thrown in for good measure. The AI then fills in any gaps, constructing what amounts to a disturbingly complete picture of your professional and personal life.

One reported example that made me put my coffee down: a user searched for a financial question and Gemini recommended consulting their CPA — then listed their actual accountant by name, having inferred the relationship from a single calendar invite. The system didn't ask. It just... knew.

(I immediately checked what my Google Calendar says about me and honestly I'd rather not discuss the results.)

For AEO practitioners, the implication is this: the AI is no longer just matching your content to a query. It's matching your content to a person. Two users typing identical words into identical search boxes at the same time will get different answers, drawn from different sources, because the system has concluded they're different humans with different needs.

This creates what I'm calling a two-layer optimisation problem, and honestly, the second layer is the one that's going to keep this industry arguing in conference Q&As for the next three years.

Layer one is the one we already understand reasonably well: making your content retrievable, authoritative, and structured in a way that AI systems can parse and cite. Structured data, topical depth, editorial credibility, earned media coverage — all the signals that determine whether your content enters the model's context window in the first place. Nothing about personalisation changes this. If the AI can't find you, it can't recommend you, no matter how perfectly suited you are to the user. You can't win a race you didn't enter.

Layer two is new, and it's the one that should be making you uncomfortable: whether your content, your brand signals, and your digital footprint align with the contextual profile of the person the AI is trying to serve. The AI isn't just asking "is this content relevant to this query?" It's asking "is this content relevant to this query, asked by this specific person, with this browsing history, these purchase patterns, and these calendar invitations?"

Think of it as a filter applied after retrieval. The AI gathers a set of potentially relevant sources (layer one), then selects from that set based on what it knows about the user (layer two). Your content can pass the first filter with flying colours and still never surface because the personalisation layer decided you're not a match for the person behind the keyboard.

The practical consequence? AEO is no longer just about being the best answer to a question. It's about being the best answer to a question asked by a specific type of person. And that requires a depth of customer understanding that most optimisation strategies — historically organised around keywords, not human beings — are completely unequipped to deliver.

The measurement implications here are, and I don't say this lightly, kind of devastating for how most of us report on search performance.

If every user gets a different answer based on their personal context, then tracking your visibility for a keyword — the foundation of SEO and AEO reporting for two decades — becomes an increasingly unreliable indicator of actual performance. A brand can hold a strong position in every keyword-tracking tool and still see no traffic, because the personalisation layer is routing users elsewhere. The inverse is equally true: a brand with no visible keyword rankings can drive significant revenue from AI search because the system is matching them to relevant users based on contextual fit rather than keyword position.

This isn't a subtle shift. It undermines the entire edifice of rank tracking as a meaningful performance metric. The question is no longer "where do we rank?" but "are the right people finding us, and what do they do when they get there?"

The metrics that matter in a personalised AI search environment are downstream of visibility: revenue from search, qualified leads, customer lifetime value. Impressions, positions, aggregate organic clicks — these become proxies that may or may not correlate with outcomes that actually matter. It's like checking the temperature outside by looking at what other people are wearing. Sometimes accurate. Sometimes you end up in a blizzard in shorts.

Google isn't the only player building personalisation into AI search, but it's the only one with the data to do it at this depth. Two decades of behavioural history across billions of users — search, email, calendar, maps, browser activity, mobile OS data — gives Google an advantage no competitor can replicate in the near term. ChatGPT, Perplexity, Claude will build their own personalisation layers, but none currently have this breadth of cross-product behavioural data. It's the digital equivalent of showing up to a poker game where one player has been reading everyone else's diary for twenty years.

The direct consequence: optimising for Google's AI surfaces is now a categorically different problem than optimising for other AI search products. The same content, targeting the same query, may perform differently across platforms — not because of quality differences, but because one platform knows who the user is and the other doesn't. A universal AEO playbook applied uniformly across all platforms is a strategy that optimises for none of them. (I know. I've seen the look on clients' faces when you tell them that.)

Right, let's get practical. Because if I've learned anything from writing this newsletter, it's that analysis without action points is just anxiety with better formatting.

Expand your brand's presence across Google's ecosystem, not just its search index. If the personalisation layer pulls signals from Gmail, Calendar, Maps, and Chrome, then every touchpoint a customer has with your brand across those surfaces becomes an AEO input. Email receipts landing in Gmail, Google Business Profile interactions, Maps visits, calendar bookings — these aren't just CRM or local SEO concerns anymore. They're feeding the personalisation engine that determines whether the AI surfaces your brand to a given user. If your business exists only as a website, you're invisible to layer two.

Shift measurement from keyword visibility to cohort-level revenue attribution. Stop asking "do we rank for this keyword?" and start asking "is revenue from AI-mediated search growing, and is it coming from the right customer segments?" GA4, server-side attribution, and CRM-linked conversion tracking can get you closer. The tooling isn't purpose-built for personalised search yet, but the direction is clear.

Invest in understanding your customer's demonstrated behaviour, not just their inferred search intent. The AI knows what your target customer has bought, read, booked, and visited. Your strategy needs equivalent depth — purchase patterns, repeat behaviours, cross-category interests, the content they actually engage with versus the content you assume they want.

Treat retention signals as discovery signals. If the AI reinforces brands a customer has already engaged with — and the flag-based architecture strongly suggests it does — then repeat purchase, email engagement, and loyalty programme participation aren't just retention metrics. They're feeding the discovery loop. Every positive interaction strengthens the signal that tells the personalisation layer to surface you again next time. The personalisation flywheel favours incumbents. That's an advantage if you're the established brand; it's a strategic problem if you're the challenger.

Accept that AEO is now platform-specific. Understand which AI ecosystem your target customers primarily inhabit and weight your investment accordingly. One-size-fits-all is over.

The existing AEO fundamentals aren't irrelevant — structured data, topical authority, earned media, editorial credibility remain the prerequisite for entering the AI's consideration set. Personalisation is a filter applied after retrieval, not a replacement for it. If your content isn't in the index, no amount of customer alignment will help.

But AEO just grew a second layer, and it demands we stop optimising for queries and start optimising for the people behind them. The keyword was always a proxy. As personalisation makes that proxy increasingly unreliable, the discipline needs to shift its centre of gravity from the question to the questioner.

We've argued in this newsletter that AEO is fundamentally SEO with a new label. That still holds for the retrieval layer. But personalisation adds a genuinely new dimension — and pretending it doesn't exist won't stop it from quietly deciding whether your content ever sees daylight.

Optimising for the query got you into the consideration set. Optimising for the person behind it is what keeps you there.

Jo Lambadjieva is an entrepreneur and AI expert in the e-commerce industry. She is the founder and CEO of Amazing Wave, an agency specializing in AI-driven solutions for e-commerce businesses. With over 13 years of experience in digital marketing, agency work, and e-commerce, Joanna has established herself as a thought leader in integrating AI technologies for business growth.

😍It was great, all I needed

🙂Good, not great

🙄It sucked