Stop Chasing “AI Visibility” and Start Running an AI-Era Traffic P&L

The real shift: your traffic is being intermediated, not just “optimized”

Scan those headlines and a pattern jumps out: everyone is obsessing over AI search, schema, “AI visibility,” TikTok sales, retail media, and new analytics channels for “AI traffic.”

Underneath all of that is one issue that actually matters to operators:
your customer acquisition is being intermediated by AI systems that sit between you and demand.

Search results, TikTok’s For You feed, retail media networks, AI assistants, shopping agents, “AI Overviews” – they are all doing the same thing:

• Aggregating demand
• Deciding who gets surfaced, cited, or linked
• Taxing that access with ad products, data capture, or both

That’s not an SEO problem. It’s not a “best time to post” problem. It’s a traffic P&L problem.

If you keep treating AI search and new intermediaries as channel quirks, you’ll keep doing busywork (schema, title rewrites, posting calendars) while your economics quietly erode.

Why “AI visibility” is three different problems, not one

One of the better recent takes framed AI visibility as three layers. For operators, that’s the right starting point – but we need to translate it into something you can run like a P&L.

Think in three layers:

1. Discovery layer – How people find a “way in” to your category and brand (search, TikTok, YouTube, retail media, AI assistants).
2. Decision layer – How they evaluate options (AI summaries, reviews, creator content, comparison tools, shopping agents).
3. Conversion layer – Where the transaction and data actually live (your site/app, marketplaces, retail media, social commerce, agents).

AI and new intermediaries are attacking all three:

• Discovery: AI Overviews and TikTok’s feed can satisfy intent without a click.
• Decision: AI answers compress your carefully built content into a bullet in someone else’s interface.
• Conversion: retail media, social shops, and agents keep the transaction – and the data – off your properties.

That’s why schema experiments show “AI citations barely moved.” You’re optimizing the wrong layer. Schema is a tactical input into a strategic shift: the platforms want to answer, not refer.

Stop asking “How do we rank?” and start asking “Where do we get paid?”

CMOs and performance leaders should reframe their core question:

Not: How do we rank in AI search / TikTok / retail media?
But: Where do we earn margin and data in a world where AI and platforms sit in front of the customer?

That means treating your acquisition like a portfolio with a real P&L:

• Some channels will be high-margin but volatile (organic search, organic social, earned creator content).
• Some will be low-margin but scalable (retail media, marketplaces, programmatic, AI agents that charge a toll).
• Some will be low-scale but strategically critical (owned community, email, direct type-in, app usage).

Your job is not to “win AI visibility.” Your job is to:

• Quantify the tax each intermediary charges (in money and data).
• Decide how much of your P&L you’re willing to hand to them.
• Invest in assets and motions that reduce that dependency over time.

A simple traffic P&L model you can actually run

You don’t need a new framework deck. You need a spreadsheet you look at monthly.

For each major source or “intermediary,” track:

• Gross traffic (sessions, calls, store visits – whatever matters).
• Net owned traffic – traffic where you own the customer relationship (email captured, app installed, account created, opt-in achieved).
• Blended CAC – including content, media, tech, and team costs.
• Intermediary tax – percent of revenue or margin captured by the platform (ad fees, commissions, data loss).
• Payback period – how long until you recoup acquisition costs.
• Incrementality – directional view, even if imperfect (geo tests, holdouts, MMM-lite).

Then classify each channel:

• Feeder – high-volume, often high-tax channels that introduce you to new customers (TikTok, retail media, marketplaces, AI assistants).
• Farmer – channels where you grow value from known users (email, app, community, loyalty, direct visits).
• Hybrid – search, YouTube, some social, where you can do both.

Your job is to ensure feeders are economically sustainable, and that you’re aggressively moving people into farmer channels where the tax drops and the LTV rises.

What this means for SEO and content: less “ranking,” more routing

The SEO headlines tell the story: algorithm updates, schema, cannibalization, title rewrites, AI KPIs. Most teams are still playing a 2015 game.

In an AI-intermediated world, SEO’s job changes:

1. From “rank for everything” to “own the right intents”

Stop chasing every keyword Ubersuggest spits out. Instead:

• Map intents into three buckets: Category education, Problem framing, Solution selection.
• Decide which ones you must own because they drive high-intent, high-margin customers.
• Accept that AI will answer a lot of generic questions and build for the moments where a human still wants depth, proof, or a tool.

2. From “get the click” to “get the relationship”

If AI search and feeds compress clicks, the clicks you do get matter more. Your SEO and content should be designed to:

• Capture an email or account with a real value exchange (tool, template, calculator, quiz, configurator).
• Drive to a high-intent action within the first session (demo request, sample, store visit, add-to-cart).
• Route people into owned surfaces (app download, community, newsletter that actually solves a problem).

3. From “publish more” to “maintain a smaller, stronger corpus”

AI systems train on your entire footprint. A bloated, cannibalized site confuses both humans and models.

Make “content pruning” and consolidation a quarterly ritual:

• Merge thin, overlapping posts into authoritative evergreen assets.
• Kill content that doesn’t drive traffic, links, or conversions.
• Refresh and reframe high-intent pages for AI-era behavior: clear answers, structured data where it actually helps, and strong calls to action.

Media buying in the age of agents and retail media: price the dependency

Retail media is eating TV. AI agents are creeping into shopping. Programmatic is being reshaped by agents that will happily spend your budget on “optimized” outcomes you can’t fully audit.

As a media buyer or growth lead, you need to:

1. Treat every intermediary as a supplier with a margin target

For each major platform (Amazon, Walmart, TikTok, Meta, Google, retail media networks, AI agents):

• Estimate their effective “take rate” on your revenue (ad spend plus fees divided by incremental revenue).
• Set a ceiling you’re willing to accept by channel and by cohort.
• Push back with budget when they cross it. Don’t argue; reallocate.

This is how you avoid waking up in two years with 40% of your gross margin quietly flowing to retail media line items.

2. Build your own “attention stack,” not just campaigns

Platforms are giving you tools to scale attention (YouTube tools, short-form editors, creator marketplaces). Use them to build assets, not just ads:

• Short-form video that can travel across TikTok, Reels, Shorts, and your own site.
• Creator content that lives on your properties as much as theirs.
• Reusable hooks and formats that become your IP, not the platform’s.

The goal: if a platform changes its algo or ownership (hello, TikTok sale headlines), you still have creative, audiences, and narratives that can move.

3. Run “agent-aware” experiments now, not later

Stripe and Google are already talking “agentic commerce.” Alexa for Shopping is pushing AI-driven purchase flows. The direction is clear: agents will increasingly decide what gets bought.

Start small but specific:

• Test how your brand and products are recommended in current assistants (Google, Alexa, Siri, ChatGPT, Perplexity).
• Identify the prompts and intents where you want to be the default suggestion.
• Experiment with feeds and structured data that these systems ingest (product feeds, reviews, FAQs, support docs).

Don’t wait for a formal “agent optimization” playbook. By then, the tax will be set.

Measurement: stop worshipping direct traffic, instrument AI-era behavior

As AI intermediaries spread, your analytics will get weirder. Direct traffic correlations, AI channels, and “dark social” will all blur attribution.

Instead of chasing perfect attribution, focus on three practical moves:

1. Separate “brand demand” from “channel performance”

Track:

• Branded search volume over time (including misspellings and short forms).
• Direct + app open trends as a bundle, not as “proof” of any one channel.
• Survey-based “how did you first hear about us?” on key forms and post-purchase.

This helps you see when AI intermediaries are answering questions but still driving branded demand your way, even if clicks are messy.

2. Instrument AI and “assistant” traffic as its own class

With tools like Google Analytics adding AI Assistant channels, treat assistant/AI-origin traffic as its own segment:

• Measure bounce, depth, and conversion separately.
• Watch how it changes after major AI product updates.
• Compare its economics to other feeder channels.

If AI traffic is high-intent but low-volume, you nurture it. If it’s noisy and low-value, you deprioritize chasing it.

3. Use business-impact triage on technical and content work

Not every technical SEO fix or schema implementation deserves a sprint. Borrow from “prioritize by business impact” thinking:

• Score each initiative on revenue at risk or upside, not just traffic potential.
• Run small tests before full rollouts (e.g., schema on a subset of high-intent pages).
• Kill or pause work that doesn’t move revenue or high-value engagement, even if it makes SEO tools happy.

What to actually do in the next 90 days

To make this concrete, here’s a 90-day plan you can hand to your team.

Week 1-2: Build the traffic P&L view

• List your top 10-15 acquisition sources by spend and volume.
• For each, estimate CAC, intermediary tax, payback, and net owned traffic.
• Classify each as Feeder, Farmer, or Hybrid.

Week 3-6: Fix the biggest structural leaks

• Pick the top 3 feeder channels where you’re overpaying for under-owned customers.
• Redesign the first-session experience to push toward owned relationships (email, app, account).
• Prune and consolidate content in one key intent cluster that matters for AI and human decision-making.

Week 7-10: Run one AI visibility experiment per layer

• Discovery: Test content or creative designed specifically for AI Overviews / assistants (clear answers, structured FAQs).
• Decision: Launch or improve a comparison/quiz/tool that AI can reference but that also converts well when humans land on it.
• Conversion: Instrument and segment AI/assistant traffic in analytics; compare its economics to other sources.

Week 11-13: Rebalance budget and roadmap

• Shift 10-20% of spend from your worst-taxed feeders into your best farmer channels and high-intent hybrids.
• Update your content and technical roadmap to focus on fewer, higher-impact initiatives tied to revenue, not just rankings.
• Set quarterly thresholds for how much margin you’re willing to pay each intermediary – and stick to them.

The platforms will keep changing their features, feeds, and AI layers. Your advantage won’t come from reacting faster to every update. It will come from running your traffic like a P&L, pricing your dependencies, and building assets that survive whatever new intermediary shows up between you and your customer.