The real problem hiding in all these headlines
Scan the recent trade headlines and a pattern jumps out: everyone is optimizing for systems they no longer understand.
AI search engines that cite fewer sites. Social platforms rewriting visibility rules. AEO platforms promising “AI-era SEO.” AI tools writing content, scoring content, and deciding which content to show. Media buying platforms changing what buyers can see and how they can bid.
Underneath all of that is one high-signal issue:
performance marketing is now happening inside opaque, AI-driven systems where the old feedback loops are breaking.
For CMOs and media buyers, this isn’t a thought experiment. It’s a P&L problem:
- Your content is being evaluated by black-box models, not just humans.
- Your ads are being distributed by AI systems that don’t explain themselves.
- Your analytics stack is still pretending the world is deterministic.
The question isn’t “how do we rank in AI search?” or “how do we make AI content?” The question is:
how do we run a commercially sane growth engine when distribution, attribution, and optimization are increasingly opaque?
The three big shifts operators need to internalize
1. Distribution is moving from “search results” to “AI answers”
Headlines about ChatGPT citing fewer sites and “AI search engines to watch” are not trivia. They signal a structural change:
- Classic SEO fought for ranked, clickable results.
- AI search fights for inclusion in a synthesized answer that may never show a link.
That changes the job:
-
From keyword targeting to entity and expertise targeting.
Search engines and AI models are building knowledge graphs: brands, people, products, categories, claims. If you’re not a recognized entity with consistent signals, you’re invisible even if you “rank” today. -
From click optimization to citation and mention optimization.
Being the source that models train on and reference matters more than being result #3 vs #5. -
From SERP testing to model-aware content design.
You’re now designing for humans and for AI systems that prefer structured, precise, low-ambiguity content.
2. Media buying is becoming “ask the machine, trust the machine”
The Trade Desk is changing how advertisers buy and what they can see. Meta and Google have already done it:
- Broad targeting instead of micro-segmentation.
- Automated bidding instead of manual bid strategies.
- Algorithmic placement instead of granular control.
The platforms are telling you: “Give us your budget, your goal, and your creative. We’ll handle the rest.”
That sounds efficient. It’s also a control problem:
- You see fewer levers.
- You get more “black box” performance.
- You’re increasingly dependent on platform-side AI to find, persuade, and convert your customers.
3. Metrics haven’t caught up to the new reality
While measurement articles scream “most marketing metrics are misleading,” most dashboards still worship:
- Last-click ROAS in a multi-touch, multi-device world.
- Channel-level CPA without any view of marginal impact.
- Vanity metrics: impressions, followers, engagement rates divorced from revenue.
In an opaque, AI-shaped ecosystem, those metrics are not just noisy; they’re actively dangerous. They push you to optimize what the platform wants you to see, not what actually compounds growth.
What operators should actually do about it
You can’t reverse the opacity trend. But you can design your growth engine to be robust to it.
The playbook comes down to five moves:
- Change what you measure.
- Change how you test.
- Change how you brief AI systems (search, social, and tools).
- Change how you structure creative and content.
- Change how you negotiate control with platforms and partners.
1. Measure the system, not the channel
When the pipes are opaque, you stop pretending you can see every drop of water. You measure the flow at the ends.
For CMOs and performance leaders, that means:
-
Adopt marginal ROI as a first-class metric.
Not “what did this channel do last month?” but “what happened to total profit when we added or removed $X from this tactic?” That’s the only way to make rational budget calls in a black-box environment. -
Use incrementality testing as your truth serum.
Geo splits, holdout groups, time-based experiments. They’re slower than dashboards, but they’re honest. You don’t need them for every tweak; you need them for every big bet. -
Collapse vanity metrics into a small set of commercial KPIs.
For example:- New customer revenue and payback period.
- Blended CAC and LTV by cohort.
- Incremental contribution margin by channel cluster (not micro-campaign).
If your weekly growth meeting can’t be run from a single-page view of these, you’re still playing the platform’s game, not yours.
2. Test at the level AI actually operates
Old testing cultures obsess over micro-variables:
- “Button color A vs B.”
- “Headline variant 7 vs 9.”
- “Interest bucket X vs Y.”
Meanwhile, platform AI is making decisions at a different level:
- Who this creative is “for.”
- What outcome is being optimized (click, view, conversion, value).
- Which creative concepts consistently generate strong post-click signals.
To get signal in an opaque system:
-
Test creative concepts, not micro-tweaks.
For example, run three radically different angles:- “Status and identity” concept.
- “Risk and loss avoidance” concept.
- “Ease and speed” concept.
Let the platform AI optimize within each concept, but you judge which idea wins on downstream revenue.
-
Test offers and economics, not just messaging.
Free trial vs paid trial. Discount vs value-add. Bundle vs single SKU. AI can’t fix a weak offer; it will simply find the cheapest people to accept it. -
Test at portfolio level.
Instead of 50 tiny campaigns, run a few well-structured “macro” campaigns that give the algorithm room to learn, then compare portfolios against each other.
3. Brief AI systems like they’re junior strategists
Whether you’re:
- Feeding content into AI search ecosystems,
- Using AI writing tools, or
- Letting platform AI build audiences and placements,
the same rule applies: garbage in, optimized garbage out.
Treat every AI touchpoint as a junior strategist who is fast, literal, and blind to context unless you give it some.
In practice:
-
Define clear, stable objectives.
“Maximize ROAS” is vague. “Acquire new customers under $X CAC with a 90-day payback” is a brief. Configure goals, conversions, and value signals to match that. -
Feed clean, opinionated data.
Fix broken tracking, standardize event naming, and ruthlessly prune fake conversions (e.g., micro-events that don’t correlate with revenue). AI will optimize to whatever you label as success. -
Give AI systems structured context.
Use consistent product names, category structures, FAQs, and schema markup. AI search and recommendation engines reward clarity and consistency more than cleverness.
4. Design content and creative for “dual audiences”: humans and models
Articles about “how to design content that AI systems prefer” are circling one idea: your content now has two audiences:
- Humans, who skim, feel, and decide.
- AI systems, which parse, classify, and summarize.
You don’t need to worship the algorithm, but you do need to stop fighting it.
Principles that matter now:
-
Precision over puffery.
Models reward explicit claims, definitions, and structures. Say exactly what your product does, for whom, and in what conditions. Vague “solutions” language gets flattened. -
Evidence over adjectives.
Case studies with numbers, methodologies, and context are more likely to be cited and summarized than fluffy success stories. Think “37% increase in qualified inquiries after X changes” instead of “huge uplift in engagement.” -
Structure over sprawl.
Clear headings, logical sections, FAQs, schema, and internal linking help both humans and models. You’re building a knowledge base, not just a blog. -
Distinctive voice on top of a clean spine.
You can still be sharp, funny, or provocative. Just build that on top of a skeleton that a machine can parse: who, what, why, how, for whom, with what result.
5. Renegotiate control: what you outsource vs what you own
As AI creeps into every tool and platform, there’s a temptation to outsource everything:
- AI writes the copy.
- AI builds the audiences.
- AI sets the bids.
- AI picks the channels.
That’s how you end up with what one copywriting shop called “AI’s trust problem” – a brand voice and growth strategy that feel generic, fragile, and hard to debug.
A more defensible split:
-
Own the strategy, outsource the grunt work.
Humans decide: positioning, promise, pricing, core narratives, and what “good” looks like. AI helps: draft, expand, summarize, resize, reformat. -
Own the measurement framework, outsource the optimization.
You define: conversion events, value weights, incrementality tests, and budget guardrails. Platforms optimize within that sandbox. -
Own the creative concepts, outsource the permutations.
Your team develops 3-5 strong creative platforms. AI helps generate variations, localizations, and format-specific cuts.
The question to keep asking: “If this AI system disappeared tomorrow, what would we still know how to do?” If the answer is “not much,” you’ve given away too much control.
How to operationalize this in the next 90 days
This doesn’t need a three-year roadmap. It needs a few deliberate moves.
Step 1: Rewrite the scorecard
- Pick 3-5 commercial KPIs that matter for the next 12 months (e.g., new customer revenue, blended CAC, 90-day payback, marginal ROI by channel cluster).
- Kill or demote metrics that don’t tie directly to those (engagement, followers, CTR) from your main CMO dashboard.
- Set a cadence for one incrementality test per quarter on a major channel or tactic.
Step 2: Restructure campaigns around how AI actually works
- Consolidate fragmented campaigns into a smaller number of “macro” campaigns per objective and geography.
- Define 3-5 distinct creative concepts per key product or segment; stop obsessing over endless small tweaks.
- Align conversion tracking and value signals with your real business goals, not what’s easiest to measure.
Step 3: Make your site and content “model-readable”
- Audit your top 50-100 URLs: are they clear on who it’s for, what it does, and what proof you have?
- Add structure: FAQs, schema, consistent naming, and internal links that reflect your actual product and category strategy.
- Create or update a small set of “source of truth” pieces (category explainers, product deep dives, case studies) designed to be cited, not just clicked.
Step 4: Set AI usage rules for your team
- Decide where AI is mandatory (e.g., first-draft generation, summarizing research) and where it’s banned (e.g., final brand voice for flagship campaigns).
- Train teams to brief AI tools with clear objectives, constraints, and examples – treat prompts like creative briefs.
- Document what’s working: prompts, workflows, and failure modes. Treat AI usage like any other repeatable process, not a toy.
The platforms will keep getting more opaque. The algorithms will keep changing. Your edge will not come from guessing the next tweak. It will come from building a growth engine that assumes opacity, measures reality, and keeps humans in charge of the parts that actually matter.