The quiet shift hiding in all those headlines
Scan those headlines and a pattern jumps out: everyone is still talking about SEO, content, CTV, Instagram, conversion wins – but the subtext is that the interface between your brand and the customer is no longer the browser or the feed.
It’s the model.
Ahrefs is studying why ChatGPT cites one page over another. Search Engine Land is pushing AI-driven SEO frameworks. Marketing.com is talking about AI citation tracking, “generative engine optimization” (GEO), and AEO metrics. Entrepreneur is telling you how to get your brand into AI search results. Microsoft’s AI ad strategy is suddenly a PPC topic.
The old funnel assumed:
- Impressions → clicks → sessions → conversions
The emerging funnel often looks like:
- Model training data → model output → in-answer brand mention → off-platform conversion
Your traffic is no longer just an outcome of your media plan. It’s an input into someone else’s model. And that model is increasingly where decisions get made.
From SERP to answer: what’s actually changing
Three structural shifts matter for operators:
1. The SERP is becoming an answer engine
Google’s “Web Guide”, AI Overviews, Perplexity, ChatGPT, Claude, and every retailer’s AI assistant are all doing the same thing: compressing the click layer.
Historically, your job in search and social was to:
- Win the click
- Monetize the session
- Retain the user
Now, for a growing chunk of queries, there is no session. There is an answer. If you’re lucky, there is a brand mention or product recommendation embedded in that answer.
2. “Ranking” is turning into “being cited”
The Ahrefs study on why ChatGPT cites one page over another is a tell. The game is shifting from:
- Rank position on a page of links
to:
- Probability of being included in a synthesized answer
That’s a different optimization problem. It’s not just about keywords and backlinks; it’s about:
- Being a canonical source for a concept, category, or spec
- Having clean, structured, machine-readable data
- Being consistently referenced by other high-authority sources
3. Measurement is breaking at the edges
GEO and AEO metrics exist because traditional analytics doesn’t see:
- When a model reads your page during training or retrieval
- When your brand is recommended inside an AI answer
- When a user converts after a model-mediated recommendation with no click path
Your dashboards still show “organic search down 12%” and “direct up 8%” and you argue about attribution models, while the real shift is that a growing share of intent is being satisfied upstream, inside models you don’t control.
Why this matters to CMOs and media buyers now
This isn’t a thought experiment. It hits three places you care about: CAC, brand equity, and channel mix.
1. CAC: the new tax on lazy brands
If your brand is underrepresented in AI answers, you’ll pay more to win the same demand:
- Search and social CPCs rise as organic and “free” discovery compress
- Retail media and marketplace ads become the only way to show up in AI-assisted shopping flows
- Affiliate and influencer programs become your rented “citation layer”
In other words, if you don’t show up in the model, you’ll buy your way back in via ads built on top of that same model.
2. Brand equity: taste is the new moat
Digiday is right: in the age of AI, taste is a competitive advantage. Models flatten mediocre content. They reward:
- Distinctive POVs that others quote and link to
- Clear, specific product differentiation that’s easy to summarize
- Owned data, studies, and frameworks that become “the” reference
If your content is generic, AI will happily rewrite it and attribute the idea to whoever expressed it more sharply or earlier. That’s the “AI’s trust problem” Copyhackers is talking about from a messaging angle.
3. Channel mix: your media plan is missing a layer
Most plans still think in terms of:
- Paid: search, social, CTV, display, retail media
- Owned: site, app, CRM
- Earned: PR, social, SEO
You now need a fourth lens:
- Model: how you show up in AI search, assistants, agents, and agentic storefronts
That “model layer” is not just an SEO problem. It’s a product, PR, data, and media problem rolled into one.
Practical playbook: from SEO to “Model Optimization”
You don’t need a new three-letter acronym. You need to treat AI systems as distribution channels with their own inputs, outputs, and KPIs.
Step 1: Map where AI is already mediating your category
Don’t start with theory. Start with reality:
- List your top 50-100 high-intent queries (search, marketplace, internal site search, customer questions)
- Run them through: Google (with AI Overviews), Bing/Copilot, Perplexity, ChatGPT, Claude, major retailer assistants in your category
- Document:
- Does your brand appear?
- Are your products recommended?
- Which sources are cited?
- What attributes or claims are attached to competitors?
This is your “AI share of shelf”. Treat it like a category review in retail.
Step 2: Fix your machine-facing surface area
Most brands are still shipping sites for humans and hoping models cope. That’s optimistic. You need to be explicit:
- Structure your data
- Implement and audit schema.org markup (products, FAQs, how-tos, organization, reviews)
- Standardize attributes (sizes, ingredients, specs, compatibility, use cases)
- Expose canonical product and content feeds (APIs, XML, JSON-LD) that are crawlable and consistent
- Clarify your canonical claims
- Write simple, unambiguous descriptions of what you are for and who you are for
- Make your positioning pages the single source of truth for those claims
- Ensure those claims are repeated (accurately) across retailers, partners, and PR
- Clean up robots and access rules
- Review robots.txt and AI crawler directives in light of Google’s updated docs and EU rules
- Decide what you want models to see (and what you don’t)
Step 3: Build “citation magnets” instead of content mills
The question is no longer “How do I publish more?” It’s “What would a model reasonably cite?”
High-probability citation assets tend to be:
- Original research and benchmarks
- Clear frameworks and definitions that others adopt (think DIRHAM, AEO, etc.)
- Definitive guides that are updated and maintained, not abandoned
- Transparent methodology pages (how you measure, test, or compare)
This is where “taste” and POV matter. Models are pattern-matchers. If everyone in your category is saying the same thing, the brand that expresses it most crisply and gets referenced most often wins the citation war.
Step 4: Treat AI engines as partners and ad platforms, not just threats
Microsoft’s AI ad strategy is a preview: ads will be injected into conversational answers, shopping flows, and productivity tools. That has two implications:
- Plan for “answer-ads”
- Expect ad units that look like “recommended options” inside AI answers
- Prepare product feeds, creative, and offers that can be sliced and recombined dynamically
- Negotiate how attribution and reporting will work before you scale spend
- Experiment with AI-native surfaces
- Test AI assistants in your own properties (site, app, CRM) where you control the model and data
- Explore agentic storefronts and bots (the OpenClaw-type tools) as extensions of your media, not just CX toys
The goal is to be present in both the organic and paid layers of AI-driven experiences, the same way you manage SEO and SEM together.
Step 5: Update your KPIs for the “model layer”
You can’t manage what you don’t measure, and right now most teams are flying blind. Borrow from the GEO / AEO conversation and define a small, sharp metric set:
- AI Share of Answers (ASA)
- % of tested queries where your brand is mentioned in the AI answer
- Track by category, use case, and competitor set
- AI Citation Volume
- Count of times your domain or brand is cited across major AI engines (via manual sampling, vendor tools, or custom scrapers where allowed)
- AI-Influenced Conversions
- Survey-based: “Did you use an AI assistant in your research?”
- Channel-based: lifts in direct and brand search correlated with AI surface tests
- Canonical Asset Coverage
- % of priority topics and queries that have a clear, up-to-date “citation magnet” on your domain
These won’t be perfect. They don’t need to be. They need to be consistent enough to guide resource allocation.
What to change in your org and budget this year
This isn’t a “wait for 2030” problem. You can adjust now without blowing up your plan.
Reframe SEO and content roles
Move your SEO lead from “traffic owner” to “model surface owner”:
- Make them responsible for schema, feeds, AI crawler policy, and ASA
- Pair them with product marketing to define canonical claims and positioning pages
- Give them a budget line for research and frameworks, not just blog posts
Give media buyers a seat at the AI table
Don’t silo AI search and assistants as “SEO” or “innovation” projects:
- Have your paid search and retail media teams test AI-native ad formats early
- Negotiate data sharing and attribution rules with platforms while you still have leverage
- Align bidding strategies with ASA and AI citation insights (for example, bid more aggressively where you’re underrepresented organically in AI answers)
Budget for “model hygiene” as a recurring cost
You already accept that:
- Brand safety tools cost money
- Site performance work is never “done”
Add model hygiene to that list:
- Quarterly audits of AI answers in your category
- Regular updates to schemas, feeds, and canonical assets
- Monitoring and correcting factual errors about your brand in AI outputs where possible
The uncomfortable truth: you’re already training someone else’s funnel
Every page you publish, every product feed you ship, every PR hit you earn is not just persuading humans. It’s feeding models that will sit between you and those humans tomorrow.
You can ignore that and keep optimizing for clicks on shrinking SERPs and increasingly paywalled feeds. Or you can treat AI engines as a new class of distribution and design your media, content, and measurement around the reality that:
- Your traffic is now an input, not just an outcome.
- Your brand is now a parameter in someone else’s model.
- Your job is to make sure that parameter is set in your favor.