The real crisis isn’t AI or algorithms. It’s that you don’t own your attention model.
Scan those headlines and a pattern jumps out:
- Google keeps changing how search and bidding work.
- Amazon, Google, OpenAI are rolling out AI-driven ad products and “journey-aware” bidding.
- Social platforms pump out new tools, formats, and metrics every quarter.
- Everyone is talking about AI traffic, AI content, and AI optimization – and almost nobody is talking about actual profit.
Underneath all of this is one problem that matters more than the rest:
most brands and performance teams do not have a clear, owned model of how attention turns into money.
So they rent one from the platforms.
That worked when targeting was cheap, cookies were abundant, and algorithms were dumb but predictable. It does not work in a world of:
- AI-powered bidding and budget pacing you barely understand
- AI traffic that looks good in-platform and dies in your funnel
- Short-form video where “views” and “attention” are totally different things
- Constant shifts in search, social, and marketplace rules
If you’re a CMO, media buyer, or growth lead, your job in 2026 is not “adopt more AI.” It’s build an Attention P&L – a simple, ruthless way to measure where attention comes from, how it behaves, and what it’s worth, independent of any one platform’s story.
Why the platform’s optimization goals are now structurally misaligned with yours
The platforms’ story: “Our AI will find the right user, at the right time, with the right message, and optimize to your objective.”
The part they skip: their objective is not your profit. It’s:
- Maximizing time-on-platform
- Maximizing auction revenue
- Maximizing dependency on their black box
That misalignment is getting worse as:
- Google rolls out “journey-aware” bidding and demand-led budgeting that move your spend based on its view of the path to conversion.
- Amazon pushes AI-driven media in a fragmented TV and retail environment, with its own closed-loop attribution.
- OpenAI experiments with ads inside conversational interfaces where “clicks” and “views” mean something very different.
At the same time, we’re seeing:
- AI content flooding search and social, creating cheap impressions with thin intent.
- AI traffic that spikes sessions but doesn’t move revenue – the “ROI problem with AI traffic” that’s finally being named.
- Social metrics (views, likes, saves) that drift further away from real business outcomes.
If you accept the platform’s optimization goal as your own, you will:
- Overpay for low-intent “engagement” and “view-through” conversions.
- Underinvest in boring, compounding assets (email, brand search, direct) that don’t make the platform money.
- Lose the ability to explain why performance is moving, beyond “the algo changed.”
What an Attention P&L actually is
Think of an Attention P&L as your internal operating system for paid and organic attention. It answers three questions:
- Where does attention come from? (channels, formats, queries, audiences)
- What does it do? (behavior, depth, time, path)
- What is it worth? (profit per unit of attention, not just ROAS)
It’s not another dashboard. It’s a set of decisions and guardrails that sit above your tools:
- What you will and won’t optimize for
- Which metrics are “vanity,” which are “diagnostic,” and which are “go/no-go”
- How you treat AI-driven traffic vs. human-intent traffic
- How you allocate budget when platform signals conflict with your own data
The goal is simple: no channel, no algorithm, and no AI model gets to define success for you. They provide signals. You decide what counts.
Step 1: Separate attention quality from attention volume
Most teams still treat “more” as “better”:
- More impressions, more views, more sessions.
- More keywords, more creatives, more audiences.
In an AI-saturated environment, that’s a trap. You need a simple way to classify attention by quality tier, not just quantity.
A practical model that works across channels:
- T1: Declared intent – brand search, high-intent non-brand search, product page views, cart activity, repeat visits.
- T2: Implied intent – category content, mid-funnel queries, engaged video viewers, email openers and clickers.
- T3: Ambient attention – short-form views, quick scrolls, top-funnel impressions, AI “research” traffic.
Then, for each tier, define:
- Target CAC / ROAS band you’re willing to tolerate.
- Max budget share of your total spend.
- Primary optimization metric (e.g., T1 = profit per order, T2 = qualified lead rate, T3 = cost per meaningful interaction).
This gives you a way to say “yes” to scale without quietly filling your funnel with cheap, low-intent attention that never pays back.
Step 2: Put a price on attention using your own downstream data
Platform metrics stop at the click or the view. Your job is to connect that to:
- Lead quality and sales outcomes
- Repeat purchase behavior
- Margin, not just revenue
A workable approach that doesn’t require a PhD:
-
Define a small set of “economic events.”
- For ecommerce: first purchase, second purchase, high-value purchase, subscription start.
- For B2B/SaaS: qualified opportunity created, opportunity to pipeline, pipeline to closed-won.
-
Back-attribute those events to their first and last meaningful attention sources.
Not every touch. Just:
- First source of intent (e.g., search query, referral, short-form video view that led to site visit).
- Last source before the economic event (e.g., branded search, email click, retargeting ad).
-
Calculate “profit per 1,000 attention units” for each source.
Attention unit = impression, view, session, or engaged view, depending on channel. The key is consistency within channel.
You’re not trying to build perfect multi-touch attribution. You’re trying to answer:
“When this kind of attention shows up, and we see it again near conversion, how much money tends to follow?”
That’s your internal price of attention – and it’s often wildly different from what the platform suggests.
Step 3: Treat AI traffic as a separate species, not just “more traffic”
Several headlines point to the same issue: AI traffic behaves differently. Whether it’s:
- AI-written content pulling in low-intent search visitors.
- Chat-based interfaces (like AI search or assistants) sending “summary” traffic.
- AI-driven placements that optimize for engagement proxies.
You need to:
- Tag and segment AI-influenced traffic where possible (e.g., landing pages built for AI search, content clusters clearly written for AI queries, campaigns using AI-optimized placements).
- Compare its funnel metrics (bounce, depth, lead quality, AOV, repeat rate) against your baseline.
- Set stricter CAC / ROAS thresholds until you have proof that it behaves like human-intent traffic.
If you don’t do this, AI traffic will quietly inflate your top-of-funnel metrics and erode your unit economics. You’ll think the algorithm is “working” because sessions and impressions are up, while profit per visitor slides.
Step 4: Redefine “optimization” so your team stops chasing vanity metrics
The social and short-form video headlines tell another story: we’re drowning in metrics that feel good and pay nothing.
You can’t fix this with another “no vanity metrics” slide. You fix it by rewriting what “good” looks like for each channel:
-
Short-form video:
- Optimize for “qualified attention rate” – e.g., percentage of viewers who watch past X seconds and perform a follow-up action (profile visit, link click, save, share).
- Set a minimum “qualified attention” threshold before scaling a creative, not just a view-through rate.
-
Search and shopping:
- Group queries by intent, not just volume. Kill or quarantine cannibalizing, low-intent terms even if they “convert” cheaply in-platform.
- Use your Attention P&L to decide which journeys deserve automated bidding and which should stay on tighter manual control.
-
Social and display:
- Separate “brand build” campaigns (measured on reach and recall proxies) from “response” campaigns (measured on profit per incremental visit or lead).
- Do not mix them under one ROAS target; that’s how you end up optimizing for the wrong thing.
The rule: no optimization target is allowed unless you can explain how it moves an economic event. If your team can’t answer that in one sentence, it’s a vanity metric.
Step 5: Put guardrails around “smart” bidding and budget automation
Journey-aware bidding and demand-led budgeting are not evil. They’re just hungry. If you don’t set guardrails, they will:
- Over-allocate to cheap, low-intent placements that look efficient on paper.
- Starve newer, higher-quality sources that haven’t had time to build data.
- Exploit attribution gaps (e.g., claiming credit for conversions driven by email or organic brand demand).
Guardrails you can implement this quarter:
- Floor and ceiling bids by intent tier, so AI can’t silently drag you into bargain-basement inventory.
- Channel-level profit targets that trump platform-specific ROAS when they conflict.
- Budget caps by attention tier (e.g., T3 top-of-funnel never exceeds X% of total spend without a review).
- Regular “attribution sanity checks” where you compare platform-reported conversions to your own first/last touch economic events.
The mindset shift: you’re not turning automation on or off. You’re deciding where automation is allowed to roam and how far.
What this looks like in practice for operators
For a CMO:
- Ask for an Attention P&L review in your monthly performance meeting, not just a channel-by-channel ROAS update.
- Force a clear split between brand, demand, and experimentation budgets – and define attention quality and economic events for each.
- Stop greenlighting “AI optimization” projects without a plan for how attention will be priced and tracked.
For a performance lead or media buyer:
- Tag campaigns by intent tier and attention type (human vs. AI-influenced) in your naming conventions.
- Build a simple spreadsheet that maps attention units → economic events → profit per 1,000 attention units by source.
- Use that sheet to argue for or against platform recommendations – with numbers that matter to finance, not just to the ad account.
For a growth leader:
- Align CRM, analytics, and media around a shared definition of economic events and attention tiers.
- Audit where your current attribution and optimization rules are rewarding cheap, low-quality attention.
- Pick one channel this quarter (e.g., Google Ads or Meta) and run a “black box vs. Attention P&L” test to show the delta in profit.
The operators who win the next five years won’t be the ones who adopt the most AI, or chase every new format. They’ll be the ones who treat attention like a financial asset with a real P&L – and force every algorithm, AI model, and platform to play by those rules.