The quiet shift that breaks your funnel math
Most teams are still optimising for “blue links” while the ground has already moved to “single answers”.
Look at the headlines you skimmed this week:
- “Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts)”
- “Google’s Preferred Sources Is Now A Global SEO Signal”
- “Google Tells Developers To Build For AI Agents, Not Just Humans”
- “Taboola’s next act: an AI answer engine for publishers”
- “AEO prompt tracking for marketing teams” and “AEO Competitor Analysis: Track AI Answer Engine Rivals”
- OpenAI laying foundations for ChatGPT ads
The pattern is simple and brutal: search is being intermediated by AI answer engines that:
- Collapse ten blue links into one synthetic answer
- Prefer a small set of “trusted” sources
- Will soon carry their own ad and affiliate layers
If you own a performance P&L, this is not a thought experiment. It is a direct hit to:
- Organic acquisition (fewer clicks, more zero-click answers)
- Paid acquisition (new auction surfaces, opaque attribution)
- Brand salience (AI summarises you into a sentence or skips you entirely)
The operators who win the next three years will treat “AI answer engine optimisation” as seriously as early teams treated SEO in 2005 and paid search in 2010.
From SEO to AEO: what’s actually changing
Traditional SEO is about ranking pages in a list. Answer Engine Optimisation (AEO) is about becoming the canonical source an AI cites or silently ingests.
Three shifts matter:
1. From rankings to “preferred sources”
Google has formalised “Preferred Sources” as a global signal. Large language models like ChatGPT and Claude show similar behaviour: a small cluster of domains gets cited disproportionately often.
In practice, this means:
- Being in the “trusted set” beats incremental on-page tweaks.
- Authority is more binary: you’re either in the club or you’re background noise.
- Once a model’s retrieval system overfits to a few domains, late entrants face a steep hill.
2. From keywords to questions and tasks
Ahrefs and others are already pushing “AI keyword research” and “content engineering”. The unit of competition is no longer a keyword; it’s:
- The question a user asks (“How do I … ?”)
- The task they want done (“Plan my trip”, “Build my media plan”, “Write my RFP”)
Answer engines sit between intent and action. They:
- Parse the question
- Assemble an answer from multiple sources
- Optionally route the user to a product, app, or site
If your content only mirrors keyword lists and not real questions and workflows, you’re feeding the model training data without earning a seat in the answer.
3. From human UX to agent UX
Google is telling developers to “build for AI agents, not just humans”. Translation:
- Machines will crawl, interpret, and transact on your site more than humans do.
- Your product and pricing data will be consumed via APIs and structured feeds.
- Agents will comparison-shop on behalf of users, using your specs, reviews, and policies as inputs.
If your site is visually beautiful but structurally opaque, you’re invisible to the new gatekeepers.
The new funnel: answer engine → agent → action
For a CMO, the most useful mental model is a new three-step funnel:
- Answer engine captures the intent (“What’s the best X for Y?”).
- Agent (ChatGPT, Claude, Gemini, Siri, Alexa, in-app copilots) interprets and refines the request.
- Action is taken on a site, in an app, or inside the agent’s own UI (purchase, signup, booking).
Your current reporting mostly sees step three. The real competition is now at steps one and two:
- Are you mentioned, linked, or recommended in the answer?
- Does the agent know how to interact with your product or catalog?
- Do your incentives align with the platform’s monetisation (ads, affiliate, rev share)?
What this breaks in your current playbook
The shift to answer engines quietly invalidates several comfortable assumptions:
Assumption 1: “If we rank, we get traffic”
Zero-click answers and AI summaries mean:
- You can “rank” and still lose the click.
- Your content can power the answer without attribution.
- Brand recall shrinks to a passing mention, if that.
This erodes the ROI story for classic content marketing unless you explicitly optimise for:
- Being cited in AI answers
- Owning branded queries and navigational intent
- Driving direct and email-based demand that bypasses search entirely
Assumption 2: “Paid search is the bottom of the funnel”
With OpenAI and others preparing ad formats inside conversational interfaces, “bottom of funnel” will move:
- From search results pages to chat threads
- From explicit queries (“buy X”) to inferred intent across a conversation
- From last-click to multi-step agent journeys that your current attribution stack will misread
Expect:
- New auction surfaces (sponsored suggestions, recommended products, “featured tools”)
- Opaque placement reporting, at least initially
- More spend routed through black-box “Performance Max”-style campaigns
Assumption 3: “Brand is what humans think, not what models think”
The “Marketing CEO” narrative is surfacing for a reason. Brand is no longer just a human memory game; it is also a machine memory game.
Two uncomfortable questions:
- When an AI is asked “Who are the leaders in your category?”, do you appear, and in what position?
- When asked “What does your brand stand for?”, does the answer sound like your strategy or like generic filler?
If models have a fuzzy or outdated view of your positioning, you are donating mindshare every time a user asks a question that should be yours.
A practical AEO playbook for CMOs and performance leaders
You do not need a 40-page “AI transformation” deck. You need a focused, commercially grounded plan. Here is a practical sequence.
1. Audit your “answer presence” across major agents
Treat this like a share-of-voice study, but for AI:
- List 30-50 high-intent questions in your category (not keywords; full questions).
- Ask them in ChatGPT, Claude, Gemini, Perplexity, and any vertical tools relevant to your space.
- Track:
- Are you mentioned by name?
- Are you linked?
- Which competitors are mentioned more often?
- Which sources are repeatedly cited (media, blogs, marketplaces, review sites)?
This becomes your “AI share of answer” baseline. Review it quarterly, like you do with brand trackers.
2. Engineer content for questions, not just keywords
The Ahrefs-style “content engineering” approach is the right instinct, but most teams stop at prompt lists. Go further:
- Map your category’s top 100 questions by:
- Stage (problem awareness, solution exploration, vendor comparison, implementation)
- Persona (end user, budget owner, technical evaluator)
- Create content that:
- Directly answers the question in the first paragraph
- Uses clear headings that mirror how a model would chunk the answer
- Includes structured summaries (FAQs, bullet lists, tables) that are easy to ingest
You are writing for two audiences: the human skimmer and the retrieval model. Both like clarity and structure.
3. Become a “preferred source” by design, not by luck
You cannot buy your way into a model’s training set (yet), but you can increase your odds of being treated as a canonical source:
- Authority: Earn links and citations from domains that models already trust (industry media, high-authority blogs, standards bodies, respected newsletters).
- Consistency: Maintain a clean, coherent content footprint. Avoid dozens of thin, overlapping pages that confuse both users and models.
- Signals: Use schema markup, clear authorship, dates, and references. Models like content that looks credible and current.
Think of this as “PR for machines”: who is vouching for you in the public corpus that models ingest?
4. Make your site agent-friendly
If Google wants you to “build for AI agents”, take that literally:
- Expose clean product and content feeds (APIs, sitemaps, structured data).
- Standardise how you present pricing, specs, and policies so they can be parsed.
- Document “how to use us” in a way agents can follow:
- Clear onboarding steps
- Example workflows
- Public docs that describe your ideal use cases
The goal: an AI agent should be able to answer “How do I set this up?” or “Can this integrate with X?” using your own materials, not a random forum thread.
5. Prepare for answer-engine ad buying
With ChatGPT ads and similar formats on the horizon, performance teams should treat 2026-2027 as the experimentation window, not the “wait and see” window.
Practical steps:
- Ring-fence a small budget (1-3 percent of paid search) for emerging AI ad surfaces.
- Insist on:
- Query-level or intent-level reporting (even if obfuscated)
- Clear creative controls (how your brand is phrased in suggestions)
- At least directional attribution hooks (view-through, conversation IDs, post-click tracking)
- Test formats that:
- Appear as “suggested tools” or “recommended providers” in relevant flows
- Drive to high-intent experiences (demos, calculators, quizzes), not generic homepages
Your goal is not to scale spend on day one; it is to learn how these auctions behave before your CFO asks why your search CPAs just doubled.
6. Tie AEO to conversion, not vanity metrics
Answer engines will tempt teams into chasing “mentions” the way early social tempted teams into chasing “likes”.
Keep it commercial:
- Track brand search volume and direct traffic alongside your “AI share of answer”.
- Use post-purchase surveys to ask “Did you use ChatGPT/Claude/other AI tools while researching?” and “Did any AI recommend us?”
- Attribute uplift from AEO efforts to:
- Increases in branded queries
- Improvements in conversion rate on comparison and pricing pages
- Lower CAC in markets or segments where you gain answer share
What to do in the next 90 days
If you run marketing, media, or growth, you do not need another abstract AI strategy. You need a short, aggressive to-do list:
- Run an AI answer audit for your top 50 questions across major agents. Document where you stand.
- Pick 10 questions you must own this year. Assign owners, content, and distribution plans.
- Clean up your machine-facing basics: schema, sitemaps, product feeds, documentation.
- Set an experimentation budget for answer-engine ad formats and conversational placements.
- Align leadership: present AEO as a revenue and margin story, not a science project. Show how it protects and grows your existing search and content investments.
The blue-link era rewarded teams who could out-optimise competitors on-page and outbid them in auctions. The answer-engine era will reward teams who can make their brand the obvious, low-risk choice for both humans and machines.