Three major publishers have now built intelligence platforms. Each one is a bet on what survives when the buyer is software.
People Inc's D/Cipher bets that content is the durable signal. Future's Helix bets that audience behavior is the durable signal. Hearst's AURA IQ bets that infrastructure position is the durable signal. They can't all be right. What each one chose to build — and what it chose to leave out — reveals what each company believes about the next five years of advertising.
The Content Bet
D/Cipher starts with the page. OpenAI embeddings process every article across 40 brands into a knowledge graph — 12,000 taxonomy terms, 100 million semantic connections. The intelligence is pre-computed offline, updated continuously, and activated through Google Ad Manager. When an advertiser buys a D/Cipher campaign, they're buying proximity to meaning, not proximity to a person. No cookies, no device graphs, no behavioral profiles. Content is the signal.
This is a genuine architectural insight. In a world where third-party cookies are gone and identity graphs are fragmenting, the one thing a publisher can always verify is what's on the page. D/Cipher's pre-computed approach also solves the inference cost problem — embeddings run in batch, not at bid time. The cost structure is fixed, not per-impression.
The results back it up. D/Cipher campaigns outperform cookie-based targeting. D/Cipher+ expanded to third-party publishers in February 2026 and is now the company's fastest-growing revenue unit, driving 26% growth in licensing and other revenue. Vogel calls it a "material contributor" for 2026.
What D/Cipher doesn't have: any mechanism for an autonomous buyer to access it. The platform activates through GAM deal IDs and direct sales relationships. When a media agency deploys a buying agent through Horizon's OS Blu or routes spend through PubMatic's AgenticOS, D/Cipher's intelligence is invisible. It can't be queried. It can't negotiate. It can't describe itself to a machine.
The Audience Bet
Future plc launched Helix in March 2026. Where D/Cipher starts with content, Helix starts with behavior. It replaces Aperture, Future's previous audience platform, which had already achieved a 450% increase in addressability across 175+ specialist sites. Helix layers predictive AI on top of that first-party behavioral data. Three pillars: Precision (real-time intent mapping), Prescription (conversion-based segment recommendations — the platform suggests audiences the advertiser hadn't considered), and Performance (a CTR guarantee).
Initial testing across 20 major U.S. campaigns showed double-digit CTR lifts. The approach is audience-forward: Helix identifies who is about to buy, not what the content says. It activates through deal IDs for programmatic buyers.
The logic is different from D/Cipher's. Future's thesis is that across 175+ specialist sites — TechRadar, Tom's Hardware, Marie Claire, GamesRadar — the behavioral signal (what someone reads across the network, how they navigate, what sequences precede a purchase) is richer than any single-page content signal could be. A reader who visits three camera review sites in a week is a better advertising target than a reader who is on a page about cameras right now.
The problem is that this behavioral signal depends on scale. Future's organic traffic declined 24% in H1 FY2026. The company's profit fell 67%. Shares dropped 30%. S&P assigned a negative outlook. An audience intelligence platform trained on shrinking traffic is training on a shrinking signal. Helix is technically sophisticated, but it's running on a substrate that's eroding underneath it.
Like D/Cipher, Helix has no MCP server. No agentic interoperability. No mechanism for autonomous buyers to discover or query its intelligence programmatically.
The Infrastructure Bet
Hearst launched AURA IQ on June 10, 2026, two weeks before Cannes. It builds on AURA, the audience platform Hearst launched in June 2024, which generates 30 personas from contextual signals across Hearst's portfolio (Cosmopolitan, Esquire, Elle, Harper's Bazaar, Car and Driver).
AURA IQ added three things its competitors don't have. First, it processes advertiser RFPs — the natural language brief an agency sends describing what they want to buy — and translates them into media plans automatically. The system reads the brief, builds custom audiences from scratch, and returns a proposal. Internal testing processed dozens of RFPs across auto, retail, CPG, fashion, and beauty verticals, with 2x CTR lift.
Second, it watches what content is performing in real time across Hearst properties and turns those trends into live audience segments — not pre-computed like D/Cipher, not retrospectively trained like Helix, but continuously generated from editorial performance signals.
Third, and this is the architectural divergence that matters: AURA IQ is MCP-native. Its Model Context Protocol architecture is explicitly designed for agentic interoperability. When a buying agent from Horizon or WPP or Publicis queries available inventory, AURA IQ can describe itself — its audiences, its capabilities, its performance data — in a format the agent understands. It can negotiate. It can respond. It is addressable by machines.
Hearst's bet is that in the agentic era, the quality of your intelligence matters less than whether buyers can find it. Being excellent and invisible is the same as being absent.
The Prebid Lesson
Header bidding went through this exact sequence fifteen years ago. The first movers built the best intelligence — proprietary yield optimization, custom floor-price algorithms, sophisticated audience valuation models. They were genuinely better at pricing impressions than the open market. And then Prebid arrived and made the infrastructure layer open and interoperable, and within three years, the quality of individual intelligence implementations mattered less than whether you participated in the standard.
The publishers who resisted Prebid — who believed their proprietary intelligence was too valuable to expose through an open protocol — found themselves excluded from the largest and fastest-growing demand pools. The ones who adopted early, even with inferior intelligence, captured the volume. Infrastructure won. Intelligence was commoditized.
MCP is following the same trajectory. At Cannes 2026, nine major ad tech companies announced MCP adoption across every layer of the advertising stack. Horizon OS Blu, Adobe CX Enterprise, TripleLift, DV Neura, Scope3, Magnite, TikTok, Smartly — the buyer-side infrastructure is coalescing around a single interoperability protocol. The agentic buying agent doesn't care which publisher has the best contextual intelligence. It cares which publishers it can query.
What D/Cipher Should Do
D/Cipher's intelligence is real. The pre-computed content graph is architecturally sound, competitively differentiated, and commercially proven. But it was built for a world where the buyer is a person — a media planner at an agency who reviews a proposal, negotiates terms, and signs an insertion order. That world is not disappearing overnight, but it is contracting. Two-thirds of advertisers plan to increase agentic spend. PubMatic reports 30+ autonomous campaigns and 20+ active agents. The transition is underway.
The gap isn't the intelligence. The intelligence is arguably the strongest of the three — D/Cipher operates across 40 brands, processes at scale, and has the deepest content taxonomy in the market. The gap is the interface. D/Cipher needs an MCP server that exposes its content graph to autonomous buyers: available audiences, performance history, semantic capabilities, deal terms. Not a new intelligence layer — a translation layer. The content graph already exists. It just needs to be addressable.
The work is not massive. An MCP server wrapping existing D/Cipher APIs — audience segments, content categories, performance metrics — is a focused engineering project, not a platform rewrite. IAB Tech Lab's ARTF specification provides the containerization standards. The tooling exists. The question is organizational priority.
Hearst understood this first. That's a lead, not a verdict. D/Cipher's content graph exposed through MCP would be the most valuable sell-side intelligence surface in the market. The intelligence is the asset. The protocol is the access layer. Build the access layer.