While Everyone Chasing GPT-5, A $1T Advertising Industry Is Secretly Imploding
Emerging privacy changes broke the advertising machine that funds the free internet. The economic engine of the web is being rebuilt mid-flight in AI first world and most brands are missing it.
Everyone's doomscrolling model leaks & new meme creation ease. Meanwhile, the economic engine of the web is being rebuilt mid-flight.
Global ad spend has crossed the $1T mark and keeps tilting digital yet the plumbing underneath has changed for good. Apple throttled mobile tracking (ATT). AI Assistants are already redirecting intent before it hits classic SERPs.
The advantage now goes to companies that control their first party foundation led audience intelligence which they own (thanks to CDP era to help adopt first party data foundation criticality).
Reader Poll: Which describes your current ad situation?
My CAC keeps rising and I don't know why
I'm still building first-party data foundation but struggling to activate it
I'm completely lost in the privacy transition
Still relying on 2019 playbooks (and it shows)
The Problem (Why Your Media Plan Feels Broken Even as Spend Rises)
Signal loss: ATT cut off cheap mobile IDs; cookie alternatives are narrower. Server-side conversions matter more than pixels.
No single "cookie-pocalypse" day: Chrome shifted from planned deprecation to status-quo + user controls while Privacy Sandbox continues without a forced cutover; degradation is a squeeze, not a cliff.
Discovery is moving upstream: Adobe tracked a 1,300% surge in holiday AI-referral traffic to U.S. retail sites and ~1,200% growth into early 2025. Intent is leaking to answer surfaces.
But here's the paradox, budgets are way up:
Meta: Q2'25 revenue $47.52B; ad impressions +11%, avg. price/ad +9%
Alphabet: Search & Other $54.2B; YouTube ads $9.8B
Amazon: Advertising services $15.7B (+23% YoY)
U.S. digital market: $258.6B in 2024, +14.9% YoY
Bottom line: You're paying more in noisier auctions while getting less deterministic signal back. The fix isn't to out-bid platforms it's to out-inform them.
Discussion starter: What's the biggest 'black hole' in your attribution right now? Hit reply- I'm tracking patterns across hundreds of CMOs dealing with this.
Two Power Centers: Intent vs Identity (Both Matter)
Intent-driven - monetizes "do I want this now?"
Leaders: Google Search & YouTube, Amazon Ads, and retail media networks (Walmart Connect, Target Roundel, Instacart). You win with clean server-side conversions, structured product data, and pre-shaped seed cohorts so their algos start closer to the pin.
Identity-driven - monetizes "who am I?"
Leaders: Meta, TikTok, LinkedIn. You win with people/account graphs, global suppression, and stage-mapped creative (problem-aware → solution-aware → purchase). TikTok alone is forecast to hit ~$32.4B ads in 2025 if the U.S. remains open.
Neither is going away. The shift is who supplies the intelligence: the platform's black box or your Audience OS.
Discussion starter: Which power center is working better for you right now? Are you seeing different results across intent vs identity channels?
🧵 TL;DR for the skimmers: $1T+ ad market, new plumbing. Intent is leaking to answer engines (AI referrals +1,300%). Win by productizing audiences you own identity + signals + portability—not by chasing shiny models.
Where Perplexity & OpenAI Fit (Executive View)
Perplexity (intent): Now selling sponsored follow-up questions next to answers an "answer SERP" primitive built to be labeled and adjacent, not injected into content. U.S. tests are live.
OpenAI (intent): ChatGPT Search is GA to everyone as of Feb 5, 2025; live web answers with citations into conversation. Ad products aren't public yet, but the surface is unmistakably search-like.
Assistants (identity): Logged-in, high-frequency usage yields first-party preference graphs. Expect cohorts shaped by assistant context + login, not cookies.
Translation: As answers displace links, intent migrates to answer surfaces; as assistants get stickier, identity migrates to assistant graphs. Bring portable audience intelligence to both.
Discussion starter: Have you noticed traffic patterns changing from traditional search? Adobe's 1,300% surge in AI referrals is wild what are you seeing in your analytics?
The Solution (Audience as a Data Product)
Stop treating "audience" as saved segments locked inside your Ad or ABM tools. Treat it as governed tables with SLAs in your warehouse/CDP:
Identity: durable person/account IDs; consent ledger; match-rate telemetry
Signals → features: recency, frequency, LTV/margin, stage, content affinity, price sensitivity
Audience intelligence: rules + ML audiences with guardrails (suppression, frequency, brand safety)
Portability: one-to-many, besides 1:1 activation to Meta, Google/YouTube, Amazon/RMNs, TikTok, LinkedIn, DV360/TTD, CTV, email/SMS from the same tables
Closed-loop learning: server-side conversions + incrementality that push nightly improvements into features and seed cohorts
Key insight: Programmatic pipes aren't dying—they're starving for better signals. Programmatic will account for ~9 in 10 display dollars and ~97% of new display growth this year (~9 in 10 display dollars; ~97% of new display growth). Feed it right.
Discussion starter: What percentage of your audience data is trapped inside individual platforms versus centrally governed? Most teams are shocked by this number.
The Playbook (Five Moves That Separate Leaders from Laggards)
Seed quality > budget size
Start with outcome-positive cohorts (high-LTV repeat buyers, Stage-3 opps). Generate platform-native LALs, then constrain with fit, consent, and frequency policies.
Global suppression day one
Exclude recent converters, actives, and low-fit across all channels; reduce waste before touching creative.
Broad vs Ported always on
Pit your ported audiences against platform broad; keep the winner per segment for 6 weeks, then re-test—don't chase channel-level winners.
Server-side everything
CAPI/Enhanced Conversions; retire orphaned pixels. Treat platform CPA as directional; manage to incrementally.
Tag AI-origin traffic as a distinct source in analytics (most teams aren't tracking this yet commit to tagging it for 90 days)
Plays as code
Global suppression as prerequisite, then tiered ABM air-cover (LinkedIn + CTV/Programmatic), win-backs (Meta/TikTok + email/SMS), high-intent surges (Google/Amazon/RMNs) all as versioned JSON (owner, budget, guardrails, KPIs). Launch in hours, not weeks.
Discussion starter: Which of these tactics would have the biggest immediate impact on your current campaigns? What's stopping you from implementing it?
How to Plug Perplexity & ChatGPT Into Your Plan Now
Be answer-ready: Publish structured, source-rich product intel (specs, claims, compatibility, pricing ranges, FAQs). Assistants prefer verifiable facts with links.
Test Perplexity intent slots: Run sponsored follow-up questions on 3–5 high-intent queries where you already win in Google Ads; tag distinct UTMs; measure assist rate and incremental CPA.
Isolate assistant traffic: Adopt server-side conversions and mark AI-origin visits; Adobe's data says this stream is material and rising.
Port your best seeds: As assistant inventory opens, feed outcome-positive cohorts and keep global suppression synced to protect lift.
Your Next 90 Days (The Exact Playbook)
Weeks 1–2: Name an Audience Owner (Like a PM)
Make relevance, fitment, freshness, resonance, match rates, identities, suppression hygiene, and incrementally someone's day job.
Weeks 3–6: Stand up server-side pipelines
CAPI/Enhanced Conversions into the warehouse; fix what breaks.
Weeks 7–10: Launch three portable plays
(a) outcome-positive prospecting
(b) global suppression retargeting
(c) Tier-1 ABM air-cover
Weeks 11–12: Layer agents with guardrails
Budget allocator (lift-aware), creative router (stage-aware), drift monitor (schema/consent/match-rate).
Report what matters: Waste removed (%), match-rate delta, incremental revenue/LTV, decision latency (days → hours). Platform dashboards are inputs, not truth.
Channel-by-Channel Reality Check (What Actually Works)
Google (intent): Search & Other $54.2B and YouTube ads $9.8B respectively in Q2'25. Treat AI Overviews/answer engines as upstream research; prioritize server-side truth and structured feeds.
Amazon & RMNs (commerce intent): Advertising services revenue $15.7B last quarter (+23% YoY). Retail media is the fastest-growing rail with ~17% CAGR ('24–'28); reconcile to contribution margin/LTV, not just ROAS.
Meta (identity at scale): +11% impressions, +9% price/ad—signal quality and suppression hygiene pay immediately.
TikTok (identity + commerce): Forecast ~$32.4B in 2025 if U.S. remains open—huge reach + creative testing. Maintain policy risk contingency given ongoing regulatory uncertainty.
Discussion starter: Which channels are performing best vs worst for you right now? Are you seeing the same patterns I'm describing?
The Bottom Line
The cheap-ID decade is over. Global ad spend crossed $1T, but the infrastructure underneath has fundamentally changed. Privacy isn't going back in the box. AI assistants are redirecting 1,300% more intent. New ad primitives are emerging.
The companies that win won't have the flashiest AI demos they'll have the cleanest audience learning loop.
Your competitors are either stuck in 2019 playbooks or chasing AI shiny objects. The durable edge now is owning identity and signals, productizing audiences, and letting AI amplify a system you control across Google/YouTube, Meta, Amazon/RMNs, TikTok/LinkedIn, and the answer layers rising beside them.
Discussion starter: What would change about your media results if you could execute this playbook perfectly? What's the biggest gap between your current state and that vision?
Additional polls for follow-up engagement:
Where do you see the biggest opportunity: Intent capture vs Identity graphs vs Starting from scratch?
What's your biggest obstacle: Technical complexity vs Internal buy-in vs Budget constraints vs Don't know where to start?
Forward this to your CMO. They're probably feeling this pain right now but don't know there's a systematic way to fix it.
Building the future of audience portability/learning loop at iCustomer. Want to see how this works in practice? Reply "90-day" and I'll send the spreadsheet playbook