The pattern nobody wants to admit: we’re addicted to tactics
Scan those headlines and you see the same thing on repeat:
- “How to create a Wikipedia page…”
- “Best time to post on TikTok…”
- “8,000 title tag rewrites…”
- “Agentic AI for social media…”
- “Generative engine optimization best practices…”
- “How to integrate Agentic AI into workflows…”
Different channels. Same story. An industry sprinting from micro-tactic to micro-tactic while the ground under it is being rebuilt by AI, retail media, and a search ecosystem that no longer behaves like “10 blue links.”
CMOs and performance leaders don’t actually have a TikTok posting problem or a title tag problem. You have a system design problem:
How do you build a performance engine that:
- Still works when Google ships another core update or UCP change
- Still works when “search” happens inside AI agents, retail media, and creator feeds
- Still works when your team is using 10 different AI tools and nobody can prove ROI
The operators who win the next five years won’t be the ones with the best prompt library. They’ll be the ones who treat AI, SEO, social, and retail media as parts of one system, designed around three things: intent, incrementality, and instrumentation.
The new funnel: from keywords to queries to agents
Look at what’s actually changing in the headlines:
- “Agentic Commerce Optimization” and “Agentic AI for social media” – AI agents doing the searching, comparing, and even buying
- “Generative engine optimization” and “Content scoring tools work, but only for the first gate” – search results and AI answers are now blended
- “MoneySuperMarket partners with ChatGPT as PPC costs continue to rise” – brands routing paid acquisition through AI ecosystems
- “Walmart says AI users build 35% bigger baskets” – AI changing not just discovery, but cart composition and margin
The old mental model:
- People search keywords
- You rank or bid
- You optimize click-through and conversion
The emerging model:
- People and agents ask questions and issue tasks (“find me the best…”, “reorder my…”)
- AI systems synthesize from multiple sources (search, product feeds, reviews, first-party data)
- Platforms decide which brands appear, in what order, and how “native” the suggestion looks
If your performance strategy is still organized around channels instead of queries and tasks, you are optimizing the wrong layer of the stack.
The three-layer model of an AI-ready performance engine
To get out of the tactic hamster wheel, you need to design for three layers:
- Demand capture layer – where queries, questions, and tasks show up
- Decision layer – how platforms and agents decide what to show
- Proof layer – how you measure, attribute, and fund what actually moves revenue
1. Demand capture: stop thinking “channel,” start thinking “intent surface”
Your media plan probably still looks like: Search, Social, Display, Email, Retail Media, etc. That’s quaint.
Instead, map your world as intent surfaces:
- Explicit search: Google, YouTube, Amazon, Walmart, app stores
- Implicit search: TikTok, Instagram, Pinterest, Bluesky, creator content
- Agentic queries: ChatGPT, Claude, Gemini, in-app assistants, OS-level agents
- Owned queries: your site search, email replies, chatbot logs, customer support tickets
For each, answer four questions:
- What are the commercial intents that matter? (e.g., “best mattress for back pain,” “cheap kid-safe snacks,” “how to switch payroll providers”)
- Where do those intents actually appear today? (not where you wish they did)
- Who currently captures that intent? (competitors, marketplaces, creators, affiliates, agents)
- What is the unit of content that wins there? (short-form video, long-form review, spec sheet, comparison table, “how to” email, structured feed)
This is where most brands quietly discover:
- They have strong Google coverage but weak YouTube and TikTok coverage on the same intent
- Creators and affiliates own the “best X for Y” queries that should be their highest-margin territory
- Retail media search terms are funded by trade budgets with zero connection to performance budgets
Fixing this is not “more content.” It’s fewer, more atomic assets designed to travel:
- One definitive “best X for Y” explainer that can be sliced into: SEO page, YouTube script, TikTok hooks, email sequence, retail media copy, and AI agent-friendly FAQ
- One performance landing page framework that can localize (as all those “local landing page” guides hint at) without fragmenting testing or cannibalizing SEO
2. Decision layer: design for how AI and platforms actually choose
The headlines around “content scoring,” “Google Natural Language,” “why Google may not use a sitemap,” and “sites that recovered from core updates” all point to the same truth:
You are not optimizing for users directly. You are optimizing for systems that stand between you and users.
Those systems care about a few boring, durable things:
- Structure – clean, consistent, machine-readable data (schemas, product feeds, FAQs, specs, pricing, availability)
- Signals – engagement, satisfaction, refunds, returns, dwell time, brand and product mentions
- Consistency – not saying one thing in ads, another in feeds, another on-site
- Safety and trust – clear policies, accurate claims, non-spammy behavior
Practically, this means three workstreams:
a) Build a “machine-facing” content layer
You already have human-facing content. You now need a machine-facing sibling:
- Structured FAQs that directly answer high-intent questions in 1-2 sentences
- Product and service specs in consistent JSON or feed formats, not just pretty PDFs
- Clear mappings between problems and products (e.g., “for back pain → models A, B, C”)
- Up-to-date pricing, stock, and shipping info surfaced via feeds and APIs
This is what generative engines and agents will draw from when they recommend you (or don’t).
b) Clean up cannibalization and content sprawl
“Cannibalization” and “8,000 title tag rewrites” are symptoms of the same disease: content created for calendar quotas, not for systems.
If you have:
- Multiple pages chasing the same intent
- Near-duplicate content across regions or brands
- Dozens of unmaintained landing pages built for past campaigns
…you are feeding conflicting signals into systems that now compress and summarize your footprint.
The AI-era move is consolidation:
- Pick a single canonical “home” for each commercial intent
- Redirect or merge weaker variants
- Standardize titles, descriptions, and core messaging so machines see one clear answer per intent
c) Instrument for system feedback, not vanity metrics
Most teams still optimize for CTR, CPC, and open rates. The systems optimizing against you care about:
- Did the user bounce back and click something else?
- Did they reformulate the query because your answer wasn’t good enough?
- Did they return the product, churn, or complain?
You need to:
- Track post-click behavior as a first-class metric in media reviews (scroll depth, secondary actions, support tickets opened)
- Feed returns, cancellations, and NPS by channel back into bidding and creative decisions
- Use tools like Search Console, retail media search term reports, and AI chat logs as intent intelligence, not just SEO hygiene
3. Proof layer: rewire budgets around incrementality, not channel folklore
Rising PPC costs, retail media arms races, and AI tools everywhere have created a simple problem: everyone claims credit, and CAC quietly drifts up.
The operators who will still have jobs after the next budget review will be the ones who can answer, with evidence:
“What is the incremental revenue and margin impact of this channel, format, or AI system?”
Three practical moves:
a) Make geo and audience experiments a habit, not a Q4 stunt
Those endless “X conferences to attend” lists are a distraction from the thing that actually matters: running your own experiments.
Design simple, repeatable tests:
- Turn off branded search in matched geos or audiences and measure the real lift
- Hold out specific regions from new AI-driven campaigns or new creative systems
- Run retail media spend-up / spend-down tests on key SKUs while watching total category sales
Bake this into the calendar: one meaningful incrementality test per quarter per major channel. No exceptions.
b) Treat AI tools as media channels with a P&L
“Successfully Using AI in Business,” “How to integrate Agentic AI into workflows,” “AI video made easy” – these are all cost centers until proven otherwise.
For each AI system (creative generation, bidding, content, chat, personalization), define:
- What it replaces (hours, vendors, media waste)
- What it should improve (conversion rate, AOV, time to launch, testing velocity)
- How you will measure it (before/after baselines, split tests, panel tests)
Then give it a simple P&L:
- Cost of tool + implementation + training
- Measured savings + incremental revenue
If you can’t do this, you’re not “AI-enabled”; you’re just adding line items.
c) Align brand, performance, and retail media on one scoreboard
The Budweiser/Pepsi/Dunkin’ Super Bowl narrative and Walmart’s full-funnel ambitions point to the same shift: the wall between “brand” and “performance” is mostly a budgeting artifact.
In an AI + retail media world, three things blur:
- Brand activity directly affects how often you’re recommended by agents and algorithms
- Retail media and marketplace presence shape perceived “default” choices
- Creator and social content act as both awareness and last-click
You need a shared scoreboard across teams:
- Category share of search (across Google, Amazon, Walmart, and key marketplaces)
- Share of “best X for Y” queries and AI answers where you are mentioned
- Blended CAC and LTV by cohort, not by channel
- Profit per incremental customer, not ROAS in isolation
What to actually do in the next 90 days
If you’re a CMO, performance lead, or media buyer, here’s a concrete 90-day plan that matters more than any “best time to post” chart.
1. Run an intent surface audit
- List your top 20 commercial intents (problems, not keywords)
- For each, map where they show up: search, social, retail, agents, owned channels
- Identify who currently wins and what content format they use
- Pick 3-5 intents where you’re clearly underrepresented and commit to fixing them
2. Build your machine-facing content starter kit
- Create or clean up:
- One canonical “best X for Y” explainer per key intent
- Structured FAQs for top 50 questions support hears
- Consistent product/service spec sheets in machine-readable formats
- Implement or fix schema markup on those assets
3. Kill content sprawl that confuses systems
- Identify duplicate or near-duplicate pages targeting the same intent
- Merge and redirect; standardize titles and on-page messaging
- Document a simple rule: one canonical asset per commercial intent
4. Put one AI tool on a real P&L
- Pick the AI system you’re already using most (creative, bidding, content, or chat)
- Define the before/after baseline and the KPIs it must move
- Run a 6-8 week controlled test and review it like a media channel
5. Ship one incrementality test
- Choose a high-spend area: branded search, a big social campaign, or retail media
- Design a simple geo or audience holdout
- Measure not just revenue, but profit and downstream behavior
- Commit to a decision: scale, reshape, or cut
The tactics will keep coming: new conferences, new posting times, new AI toys, new “best practices.” Most of them will be obsolete before your next planning cycle.
An AI-ready performance engine built on intent, incrementality, and instrumentation will not.