Answer-Engine Ads Arrive — ChatGPT Self-Serve, AEO Sensor, and the SEO Budget Reset

Key Takeaways

• OpenAI opened the ChatGPT Ads Manager to self-serve buyers on May 5, dropped the original $200,000 minimum, and added cost-per-click bidding alongside the pilot's $60 CPM model — the entry-cost barrier the pilot put in place in February is now gone.
• StackAdapt joined the ChatGPT advertising pilot the same week as the self-serve launch, bringing the count of named technology partners to five alongside Criteo, Kargo, Adobe, and Pacvue — programmatic supply now reaches answer-engine demand through the same buying surface marketers already use.
• HubSpot launched AEO Sensor as a free public dashboard on May 14, tracking daily volatility plus weekly visibility, citations, and AI-referred traffic for ChatGPT, Gemini, and Perplexity — the answer-engine equivalent of the measurement instruments search marketers spent twenty years building.
• The 60/40 SEO/AEO budget split that consultancies have been proposing as the 2026 baseline is plausible at the budget level and unworkable at the measurement level — the per-engine citation behavior, query share, and referral attribution differ enough that the 40% must be subdivided across at least four different optimization disciplines.
• Anthropic remains the only top-three lab with no advertising surface in production today; whether Mythos stays ad-free through Q4 2026 is the single cleanest test of whether the answer-engine ad market is a structural feature of consumer AI or an OpenAI-specific monetization choice.

The Argument

The infrastructure for advertising inside answer engines arrived this week. Not the concept — the concept has been live since OpenAI's pilot launched on February 9, 2026, with Target, Ford, Mrs. Meyer's, and Adobe as the first buyers. What arrived this week is the entry-cost reduction that turns a pilot into a market and the measurement layer that turns a market into a budget line.

On May 5, OpenAI opened the ChatGPT Ads Manager to every US business as a self-serve product, dropped the $200,000 minimum commitment that gated the pilot, and introduced cost-per-click bidding alongside the original $60 CPM model. The same day, StackAdapt — a Toronto programmatic platform with broad mid-market reach — became the fifth named technology partner in the pilot, joining Criteo, Kargo, Adobe, and Pacvue. Nine days later, HubSpot launched AEO Sensor as a free public dashboard for answer-engine visibility, volatility, and citation data across ChatGPT, Gemini, and Perplexity.

The three changes are not separate stories. They are the supply, demand, and measurement components of an advertising stack arriving inside a single nine-day window, and they collectively transform the cost structure of brand discovery in a way that the previous decade of search-marketing optimization will not absorb without restructuring.

This article walks through what arrived, what it competes with, what the 40-percent-of-discovery number does and does not mean, how the 60/40 budget recommendation breaks down once measurement begins, which marketing categories get disrupted first, and the testable forecast for the next twelve months. The broader thesis is that answer-engine advertising is the first new top-of-funnel distribution surface to reach self-serve scale in approximately fifteen years, and the budget consequences are larger than the current discourse around "AEO best practices" suggests.

What Actually Arrived This Week

Three discrete shipments inside a nine-day window need to be separated cleanly before the analysis becomes useful.

May 5 — Self-serve Ads Manager. OpenAI announced that any US business can now run ChatGPT advertising inside the OpenAI Ads Manager interface without prior approval, without the $200,000 pilot commitment, and with the choice of CPM or CPC bidding. The Ads Manager exposes the same conversation-context targeting that pilot advertisers received, plus a custom-audience capability that lets buyers upload first-party lists and define lookalikes inside the ChatGPT inventory. The change is mechanically small — a dashboard, a billing flow, a minimum waiver — and economically enormous. Below approximately a $1,000 monthly spend the pilot was not a real market; with the minimum gone, the pilot becomes available to every long-tail advertiser in the existing programmatic ecosystem.

May 5 — StackAdapt as fifth technology partner. The partner roster matters because it determines which existing buying workflows can route demand into ChatGPT inventory without requiring an advertiser to relearn the buying interface. Adobe brings the enterprise creative pipeline; Criteo brings performance retargeting at programmatic scale with roughly 17,000 advertiser clients; Kargo brings mobile and premium publisher inventory; Pacvue brings retail-media expertise; StackAdapt brings mid-market programmatic with a demand-side platform footprint that overlaps with the long-tail of direct-response and category brands. The five together cover most of the demand surfaces a CMO already buys against. There are no remaining significant programmatic platforms unrepresented in the pilot. The implication is that ChatGPT inventory is now structurally accessible from inside the buying tools marketers already operate.

May 14 — HubSpot AEO Sensor as free public dashboard. AEO Sensor tracks three signals across ChatGPT, Gemini, and Perplexity: daily answer-engine volatility, weekly AI-referred traffic, and weekly visibility and citation data per industry. The dashboard is free and public; HubSpot's paid AEO product provides brand-specific visibility tracking and recommendations. The positioning is exactly the same logic Moz and SEMrush used in 2008 — release a visible volatility instrument as a community good while monetizing brand-level insights as a paid product. The strategic effect is that AEO now has a publicly cited industry barometer, which is what budget owners need before they will reallocate spend.

The three events together describe a complete advertising infrastructure arriving inside a single quarter. The pilot phase between February 9 and May 5 was a discovery exercise — a few dozen named brands, a heavy minimum, hand-held onboarding, no measurement layer. The post-May-5 phase is a market.

The Pilot Timeline

The compression of the pilot is worth examining because it is unusual. Programmatic platforms typically take eighteen to thirty-six months from private beta to self-serve general availability — sufficient time to develop buying interfaces, attribution standards, fraud controls, and policy review. OpenAI compressed that arc to roughly twelve weeks.

The acceleration was made possible by something search advertising never had at this stage — incumbent demand. Every named partner already operates a fully built programmatic stack: buying interface, billing, audience management, attribution. The ChatGPT integration becomes another inventory source inside an existing supply graph rather than a new advertising business that has to build buyers and creative tooling. The work OpenAI compressed was the inventory formatting and policy layer, not the demand assembly. The demand was already sitting in Criteo and Adobe and StackAdapt waiting for a new inventory source that could be reasoned about programmatically.

The implication for competitors is significant. Google Search built its advertising market over five years; Facebook built its over four; TikTok needed three. ChatGPT's advertising market reached self-serve in roughly twelve weeks because the demand-side infrastructure was already standardized. Any consumer answer engine launching with a similar partner approach can plausibly compress the arc the same way — which means Gemini, Perplexity, and whatever Anthropic eventually decides about Mythos are all on a much faster clock than search advertising's historical pace would suggest.

Why $60 CPM and Self-Serve CPC Matter Together

The pricing decisions OpenAI made for the pilot are intentionally legible. $60 CPM positions ChatGPT inventory at a premium to most programmatic display (roughly $5–15 CPM) and at a discount to high-intent search (roughly $80–200 CPM equivalent for branded queries). The implicit claim is that ChatGPT inventory is closer in value to search-intent than to display, because conversational context is treated as a stronger signal than browsing context.

CPC bidding changes the buying calculus more than the CPM number does. Under CPM, ChatGPT inventory only made sense for brands with predictable click-through rates against conversational placements — early adopters with strong creative and clear intent matches. Under CPC, the platform itself absorbs the variance in click-through performance, and the buyer pays only for qualified outcomes. That is the bidding model performance advertisers historically prefer and the one that pulled the bottom 90% of programmatic spend onto search and social a decade ago.

The combination of self-serve, no minimum, and CPC bidding is the threshold condition for what the trade press has called "long-tail entry" — the participation of brands too small to commit $200,000 to a pilot but large enough to run $2,000 monthly programmatic campaigns. Below the May 5 threshold the addressable market was a few thousand advertisers; above it the addressable market is approximately the size of the US programmatic display market, which is north of two hundred thousand active buyers. The order of magnitude change in addressable demand is what makes the May 5 change a market opening rather than a milestone announcement.

The $60 CPM positioning is interesting because it locks in a price that the platform will struggle to defend if Gemini or Perplexity enter the market with aggressive launch pricing. Search advertising's pricing power came from auction depth, not from list price; the auction surplus is what made Google's ad business durable. ChatGPT's auction is still thin enough that the listed $60 is the price most buyers actually pay. That will change as the auction deepens, but it makes the early pricing brittle to competitive entry.

The 40 Percent Number — What It Does and Doesn't Mean

The headline statistic the AEO discourse leans on is that answer engines now handle approximately 40% of information-discovery queries — across Google AI Overviews, ChatGPT Search, Perplexity, Gemini, and the long tail. The number is plausible, the number is also misleading.

The 40% figure aggregates extremely heterogeneous behaviors. Google AI Overviews are inline summaries above traditional search results — the user still sees the SERP and still clicks blue links. ChatGPT Search is a destination the user navigates to deliberately for a synthesized answer. Perplexity sits between the two with stronger citation rendering. Gemini behaves more like ChatGPT Search inside the Gemini app and more like AI Overviews inside Google. A "discovery query" routed through AI Overviews is structurally different from one routed through ChatGPT, and lumping them together produces a headline that overstates the disruption to traditional SEO and understates the disruption to high-intent commercial search.

A more useful decomposition: AI Overviews still leave traffic on the table for publishers, especially for queries where the inline summary is incomplete or where the user needs to verify. ChatGPT Search rarely produces an outbound click — the model answers the question. Perplexity falls roughly in the middle. That means the consumer-funnel disruption is highest for ChatGPT-class destinations and lowest for AI-Overviews-class augmentation. AEO strategies that assume the 40% is homogeneous misallocate effort across engines.

The implication for advertising buyers is direct. ChatGPT inventory is the strongest displacement for traditional search advertising's branded and commercial-intent buckets, because the user who used to type "best CRM for small business" into Google now asks ChatGPT and never reaches a SERP. AI Overviews inventory — which is Google's response — is a different competitive problem and one Google will solve internally. The advertising stack OpenAI just opened sits squarely against the slice of search demand that no longer hits Google at all, which is a measurable and growing subset of the funnel.

How AEO Differs from SEO Structurally

The migration from search optimization to answer-engine optimization is sometimes framed as a vocabulary shift — same content, new keywords, slightly different markup. That framing is wrong in ways that matter for budget allocation.

SEO optimizes for a stable ranking surface (Google's SERP) with predictable inputs (links, on-page signals, technical health, content quality) and a single measurable outcome (organic click-through). AEO optimizes for a fragmented citation surface across at least four engines, each with distinct retrieval behavior, citation conventions, refresh cadence, and the property that the "answer" the engine produces does not deterministically include any specific source. The optimization function is closer to multi-engine citation maximization than to ranking optimization.

Four structural differences:

Citation behavior is per-engine. Perplexity surfaces citations prominently and links generously; ChatGPT Search cites narrowly and inconsistently; Gemini surfaces citations differently inside the app versus inside AI Overviews; Claude surfaces citations only when the user explicitly invokes web search. The same content asset gets cited at radically different rates across engines. A content strategy that optimizes for ChatGPT citation does not generalize.

Volatility is daily, not weekly. SERP volatility was a weekly phenomenon because Google's index refresh schedule was predictable. Answer-engine volatility is daily because model behavior drifts as new training data lands and as system prompts change. AEO Sensor's daily volatility tracking is a direct response — the measurement cadence has to match the underlying change cadence.

Entity-led, not query-led. Search optimization is keyword-led — what phrases do users type, and what content matches those phrases. Answer-engine optimization is entity-led — what entities are users asking about, and how does each engine represent those entities in its retrieval index. The unit of optimization moves from queries to entities, and entity management becomes a new discipline most marketing teams do not staff.

Attribution is broken. A user who learns about a brand from ChatGPT and then clicks through to a paid search ad shows up in Google Ads attribution as a paid-search-driven conversion. The answer-engine touchpoint is invisible to the existing measurement stack. AEO Sensor's "AI-referred traffic" metric is a workaround, not a solution — the genuine cross-engine attribution problem will take several years to fix and is the largest measurement gap in the 2026 marketing stack.

The Budget Reallocation Math

Consultancies have converged on a 60/40 SEO/AEO budget split as the 2026 baseline recommendation. The recommendation is roughly correct at the headline level — answer engines now handle enough discovery volume that allocating 40% of search-and-discovery spend toward AEO is defensible. But the decomposition of that 40% is where the recommendation breaks down.

A working decomposition of the 40% AEO allocation, sized to actual workload:

• Approximately 12 points toward entity and structured-data work — schema markup, entity disambiguation, knowledge-graph alignment. This is the closest analog to traditional technical SEO and the easiest reallocation from existing teams.
• Approximately 10 points toward content engineering for answer-engine retrieval — long-form authoritative content with clean attribution structure, FAQ-formatted summaries, comparison content, citation-friendly paragraph structures.
• Approximately 8 points toward measurement and tooling — AEO Sensor subscription if applicable, per-engine citation tracking, internal attribution work, the new measurement layer.
• Approximately 6 points toward ChatGPT advertising inventory — the new paid surface that did not exist three months ago.
• Approximately 4 points toward distribution-side investments — Wikipedia presence, authoritative reference site placement, the slow-burn entity authority that answer engines weight heavily.

The composition is more important than the headline split because the five sub-buckets compete for different internal capacity. Entity work needs technical SEO skills the team already has; content engineering needs editorial capacity that is usually thin; measurement work needs analytics capacity that is almost always thin; paid ChatGPT inventory needs programmatic-buying skills that often sit in a different team entirely. The 60/40 number understates the operational complexity.

What Gets Disrupted First

Four marketing categories absorb most of the answer-engine disruption.

Mid-funnel SEO content marketing. "Top 10 best X" comparison articles, buying guides, evergreen long-form — the genre that consumed the largest share of SEO content investment over the past decade — is exactly what ChatGPT and Perplexity answer most efficiently. The content investment still has value as an authority signal to the engines, but the direct-traffic outcome it historically produced collapses. Teams that ran content marketing as a direct acquisition channel rather than as a brand-authority signal will see the direct attribution disappear first.

Performance SEM at the long-tail. Branded and high-intent queries continue to flow through search and continue to convert. The long tail of mid-intent informational queries — the queries SEM has historically had moderate performance on — increasingly route through answer engines. Performance marketers will see SEM efficiency hold on the head and degrade on the tail, which looks like a gentle decline in aggregate ROAS rather than a sudden collapse.

Agency models built on hourly SEO retainers. The skill mix needed for AEO is different enough from traditional SEO that agencies built around keyword-and-content workflows have to retool. Larger agencies will rebuild service lines; smaller specialist agencies that focus on technical SEO will likely be acquired into AEO consultancies through 2026 and 2027.

Affiliate and lead-gen content sites. Sites whose business model is "rank for a buying-intent query and route traffic to advertisers" face the sharpest disruption. The answer engine replaces the entire navigational step, and the affiliate revenue collapses with it. This category has been visibly under stress since AI Overviews launched in 2024; the May 5 advertising infrastructure expansion accelerates the existing pressure.

What Won't Be Disrupted (Yet)

Three categories are largely insulated through at least the first half of 2027:

Transactional commerce intent. Users still type "buy [specific product]" into Google or directly into Amazon, and the conversion path goes through established product search. Answer engines summarize comparison content but do not yet handle transactional product flow at scale.

Local intent. "Plumber near me," "dentist in [zip]," "best Thai restaurant in [neighborhood]" — local-intent queries continue to route through Google Maps and local SERPs. Answer engines have weak local intent handling and weaker local conversion paths.

Brand demand. Users who type a brand name are still navigating to the brand. Brand-defense search advertising remains intact. Answer engines sometimes summarize brand information, but the user already knows what they want.

The pattern is consistent: answer engines disrupt informational and mid-funnel queries decisively, and disrupt transactional and local queries slowly. The Q4 2026 budget shift will likely concentrate the AEO investment on the informational and mid-funnel half of the funnel while keeping transactional SEM and local SEO funded at near-prior levels.

The Measurement Crisis AEO Sensor Tries to Solve

The single hardest problem in AEO is measurement, and HubSpot's May 14 launch is the first public attempt at the equivalent of the SEMrush "Sensor" volatility instrument from 2014.

AEO Sensor tracks three things: daily answer-engine volatility (how much answers are shifting day-over-day across ChatGPT, Gemini, and Perplexity), weekly AI-referred traffic (how much site traffic is arriving with referrer signatures from answer engines), and weekly visibility and citation data per industry (which brands and entities are being cited where).

The dashboard does three useful things and one structurally limited thing:

It provides a publicly visible volatility instrument that budget owners can point to when arguing for AEO allocation. That is the same function the old SERP-volatility dashboards served for search. Marketing leaders need a publicly cited "the market is moving" signal to justify reallocation internally.

It establishes a vocabulary — visibility, citations, volatility, referral traffic — that has not had a canonical reference point in AEO discussion. A shared vocabulary is a precondition for cross-team coordination.

It anchors industry-level baselines that brand-specific tools (HubSpot AEO, which is paid) can compare against. The free-paid tier split is mature commercial strategy.

The structural limit is that AEO Sensor does not solve the cross-engine attribution problem, because that problem is unsolvable from outside the engines themselves. The engines do not consistently emit referrer headers, do not provide unified click reporting, and frequently produce zero outbound clicks at all. The "AI-referred traffic" metric is an approximation based on self-reported user-agent strings and observed traffic patterns. The number is useful as a directional indicator. It will not be reliable enough for finance-grade attribution through at least 2027.

The Agency Pricing Model Reset

The hourly retainer that defined SEO agency economics for fifteen years does not survive the AEO transition cleanly, and the agency-side restructure is the second-order effect most marketers underweight when they plan their own budget shift.

Traditional SEO retainers monetized three things — keyword research, content production, and technical site work — and priced against the predictable cycle of monthly ranking reports and content output. The work was labor-intensive, the cadence was monthly, and the deliverable surface was concrete enough that client and agency could agree on what success looked like. The retainer model worked because all three sides — client, agency, and Google's algorithm — were in rough equilibrium about what optimization meant.

AEO breaks the equilibrium in three ways. First, the optimization surface is four-to-five engines rather than one, and the per-engine work does not share labor inputs cleanly. Citation optimization for Perplexity is not sufficient for ChatGPT Search and not relevant to AI Overviews. Second, the deliverable cadence collapses from monthly to daily — answer-engine volatility makes the monthly report less informative than a daily volatility trend. Third, the measurement layer no longer terminates in click-through; the value an agency delivers is increasingly invisible at the click level and visible only at the citation level, which most reporting tools do not expose well.

The agencies that survive the transition are restructuring around three new pricing axes: outcome-based pricing tied to citation share rather than ranking position, retainer plus performance hybrid models that align with the longer feedback loop, and discipline-specific specialization where a single agency handles entity work or content engineering rather than the full stack. The boutique full-service SEO agency model is the segment most at risk. Larger holding-company agencies will rebuild service lines around AEO disciplines and use scale to absorb the measurement-tooling investment the smaller agencies cannot fund.

The downstream consequence for client teams is that the agency procurement process most companies use — RFP for SEO services, evaluate proposals, select on price and case studies — is calibrated for a discipline that no longer reflects the work. Procurement teams will have to learn the new discipline vocabulary before they can evaluate the new agency landscape, and that learning curve adds friction to the budget reallocation that the rest of this analysis assumes will happen on a quarterly cadence.

The Anthropic and Google Question

The cleanest test of whether answer-engine advertising is a structural feature of consumer AI or an OpenAI-specific monetization choice is whether Anthropic's Mythos and Daybreak positioning stays ad-free through Q4 2026. As of mid-May, Anthropic has not announced any advertising surface in any consumer product. The pricing for Claude consumer tiers (Pro and Max) suggests a subscription-only monetization path. The enterprise product is the Anthropic business model in a way that is not true of ChatGPT's consumer business.

Google's position is intermediate. AI Overviews monetize through existing search advertising — the ad units appear alongside the inline summary the same way they appeared alongside organic results. The destination Gemini app has not yet introduced an advertising surface as of mid-May, but the infrastructure for one is trivially adjacent to existing Google Ads inventory. Whether Gemini-app advertising launches in 2026 is partly a function of how the ChatGPT pilot performs commercially through the next two quarters.

The interesting third position is Perplexity, which has been monetizing through sponsored citations in a limited way since late 2024 and through its enterprise product. Perplexity's advertising surface is not yet at the self-serve depth ChatGPT just reached, but the company's positioning has been the clearest about treating answer-engine advertising as a structural part of the business from the start.

The three labs and the search incumbent thus span the full strategic space: ChatGPT goes broad consumer-advertising; Gemini retains search-style inventory inside Google and may or may not extend to the Gemini app; Perplexity monetizes sponsored citations narrowly; Anthropic stays subscription-only. The market will reveal which positioning is durable, and the answer matters for budget planning. Marketers planning 2027 cannot assume the four engines they need to optimize against will have similar ad surfaces — they almost certainly will not.

The Next Twelve Months — Testable Claims

A handful of falsifiable claims for the period from May 2026 through May 2027 that the answer-engine ad stack now makes worth tracking:

• ChatGPT ad inventory revenue will exceed $1.5 billion run-rate by Q1 2027. The pilot pricing at $60 CPM combined with the self-serve expansion gives OpenAI a path to a multi-billion advertising business inside one fiscal year. The directional question is whether the auction deepens fast enough to drive effective CPM higher; the magnitude question is whether usage volume grows fast enough to fill the inventory.

• At least one additional consumer answer engine (Gemini app or Perplexity) will open a self-serve advertising surface before Q4 2026. The competitive pressure from ChatGPT's market entry is large enough that competing engines cannot afford a twelve-month gap in monetization. The exception is Anthropic, which we expect to remain subscription-only through the period.

• Anthropic announces a subscription-only commitment for Claude consumer products through 2026, treating the absence of ads as a positioning differentiator the same way Apple did for the App Store privacy posture. This is consistent with the broader Mythos positioning and with the competitive framing of the services-disintermediation thesis.

• The 60/40 SEO/AEO budget split widens to 55/45 by year-end 2026 for mid-to-large brands and to 50/50 for enterprise brands with high informational-query funnels. The 40% allocation is a starting point; measurement instruments will reveal larger answer-engine traffic shares faster than the budget reallocation has so far kept pace with.

• AEO Sensor reaches the same brand-recognition position SEMrush's Sensor reached in 2016 — the publicly cited volatility instrument the industry references in trade press and conference keynotes. The function it serves is timing — when an industry needs a barometer, the first credible barometer becomes the canonical one. HubSpot moved early enough to claim that position.

The agentic-frontier token pricing floor prediction is adjacent here for a non-obvious reason. Answer-engine advertising introduces a revenue stream that meaningfully reduces the pressure on token pricing as the primary monetization vector — a consumer answer engine generating $60-CPM advertising revenue against billions of conversational impressions has less need to compress per-token economics aggressively. The ad market makes the pricing floor easier to defend, not harder.

What This Means for Marketing Teams Today

The actionable framing for marketing leaders heading into Q3 2026 planning is not "should we do AEO" — that question is settled. The actionable framings are:

What is the per-engine optimization plan? ChatGPT, Gemini, Perplexity, Claude (when relevant), and AI Overviews require distinct citation and visibility approaches. The single biggest mistake is treating "AEO" as a unified discipline. It is four-to-five disciplines that share a vocabulary.

Who owns answer-engine paid media? ChatGPT advertising is now buyable through the same DSPs the team uses for display, but the optimization mental model is closer to high-intent search than to display. The internal ownership question — does this live with the SEM team, the programmatic team, or a new function — needs an explicit answer this quarter.

What is the measurement maturity ramp? AEO Sensor provides industry baselines but not brand-level attribution. The internal measurement build that closes the gap is a six-to-twelve-month effort and the planning needs to start now.

What is the budget reallocation pace? The 60/40 target is reasonable as a 2026 endpoint, but the migration cadence matters more than the destination. Reallocating too fast strands existing SEO infrastructure; reallocating too slowly cedes AEO position to faster-moving competitors. A quarterly reallocation cadence of 3-5 points per quarter is what most well-resourced teams will execute.

The advertising and discovery layer of the marketing stack is more contested in mid-2026 than at any point since the rise of social media buying in the early 2010s. The infrastructure that arrived between May 5 and May 14 is the opening of that contest, not the conclusion. The next twelve months will reset the budget composition that has been roughly stable for a decade, and the operational work to handle that reset is the actual hard problem most marketing teams will spend Q3 and Q4 2026 trying to solve.

Further Reading

• The services-disintermediation analysis covers the parallel consolidation on the enterprise-services side that this consumer-advertising shift mirrors on the consumer side.
• The OpenAI Daybreak / Claude Mythos counterpoint walks through the competitive positioning context that makes Anthropic's ad-free posture a deliberate strategic choice rather than a default.
• The Cloudflare 1,100 layoffs analysis is the workforce-restructuring counterpart to the marketing-budget restructuring this article describes — the same AI-platform pressure applied to a different cost center.
• The Colossus 1 compute-rent thesis is the supply-side companion to the demand-side market opening described here — the inference economics that make consumer answer engines economically viable in the first place.
• The agentic-frontier token pricing floor prediction is the testable thesis the new advertising revenue stream reinforces rather than threatens.