The Cannes Reckoning - what the industry got right, what it missed, and what Q3 demands

THE CUTTING EDGE

AI Brand Strategy   ·   June 2026   ·   Tim de Rosen, CEO AIVO

Cannes Lions 2026 is over. The panels have packed up. The rosé has been cleared. And the marketing industry has arrived at a broad consensus: AI search is real, AI recommendation matters, and brands that do not show up in AI answers are invisible.

That consensus is correct. It is also incomplete in a way that will become commercially consequential before the year is out.

This edition of The Cutting Edge is a debrief on what Cannes got right, what it missed, and what the next six months actually require of CMOs and CDMOs who take this seriously.

PART I · WHAT CANNES GOT RIGHT

The visibility consensus is real and earned

The volume of serious research published at Cannes this year was significant. Havas published The Science of Desire — 87,500 respondents, 2,400 brands, 10 markets — finding that desirable brands are up to 4x more likely to be cited by AI. Conductor published intent-type analysis across 14,000 API calls. Gartner predicted that earned media budgets will double by 2027 because AI relies on earned, not paid, media.

The collective message from Cannes was accurate: brands need to be present in AI systems, structurally readable, and consistent across the sources AI learns from. This is a Layer 2 argument — by which we mean the infrastructure layer that makes brands readable and retrievable by AI systems at the discovery and first-prompt layer — and it is the right argument for where most brands currently sit.

The shelf has moved inside the model. You cannot buy your way onto it. You earn your way in.

The Kraft Heinz CMO put it well at Cannes: just because your message was delivered does not mean it was received. Paid media does not buy AI recommendation. Earned authority does. That is a fundamental shift in how marketing budget creates brand outcomes, and Cannes was right to make it the dominant theme.

PART II · WHAT CANNES MISSED

Citation is not recommendation. Nobody measured the gap.

Here is what the Cannes panels did not discuss: what happens after the brand is cited.

The entire week's conversation was organised around a single question: does the AI know your brand? The measurement frameworks presented — share of voice, citation rate, mention frequency — are all answers to that question. They measure whether the brand is present in the AI's awareness at the first turn of a conversation. They do not measure what happens at the decision turn.

AIVO has been running structured probes across multi-turn AI buying journeys for two years. The finding is consistent across 12,000+ journeys spanning ten industries:

87.3% of brands present at turn one are displaced by a competitor before the final recommendation at turn four

75.7% of facts the model possessed about a brand were not deployed when it made an actual purchase recommendation

95.7% of brands are recognised by the model at turn one — the knowledge gap has largely already closed

The third number is the most important. The model already knows your brand. The gap is not between what the model knows and what you want it to know. The gap is between what the model knows and what it uses at the moment it makes a recommendation. We call this the Linkage Gap.

Three independent research programmes published since Cannes confirm it. Jack et al. (arXiv, May 2026) found that L1 category leaders appear in nearly every relevant retrieval but win only 25-41% of recommendation slots — across 37,000 runs. Conductor's Purchase intent finding is the most commercially alarming: AI recommends different brands to someone ready to buy more than half the time. In practice, that means a brand that appears in AI purchase recommendations 40% of the time is losing 60% of potential AI-mediated purchase moments — invisibly, without any signal in current dashboards. Havas found that desirable brands are 4x more likely to be cited, without addressing whether that citation survives to the recommendation.

Desire gets you cited. Activation gets you chosen. Cannes talked about the first. Nobody talked about the second.

The measurement infrastructure for visibility — share of voice, citation rate, brand overlap — is now well-developed and well-funded. The measurement infrastructure for what happens at the decision turn does not yet exist at most enterprise brands. That is the gap Q3 and Q4 will begin to close.

PART III · THE ARCHITECTURE THE INDUSTRY NEEDS

Three axes. Most brands are operating on one.

The Beyond Visibility working paper (Sheals and de Rosen, July 2026) sets out the corrected architecture. It distinguishes three positions:

The Possession Axis

Training-data partnerships, knowledge graphs, in-LLM applications. Shapes what the model knows. Our data shows 95.7% of brands are already recognised at turn one without any such investment. The knowledge gap has largely closed. Further possession-side investment addresses a problem that has already been largely solved.

Layer 2 — Mention (GEO)

Schema markup, AI-readable site standards, citation engineering, entity SEO, earned media. Shapes where the brand is mentioned at first prompt. The territory of every funded AI visibility platform and the subject of most Cannes panels. Necessary. Not sufficient.

Layer 3 — Activation

The infrastructure that surfaces the right brand fact, in the right structured form, at the moment a model is forming a recommendation. The layer nobody has built, named, or invested in proportionate to its importance. The layer where the Linkage Gap lives. And where the next era of AI-native brand competition will be decided.

The structural test is simple: does the output appear in the model's first-prompt answer, or in the model's recommendation at the decision turn? The former is Layer 2. The latter is Layer 3.

PART IV · THE NEXT BIG THING

The Human Agentic Gap: what Q3 and Q4 will force brands to confront

The Linkage Gap is the immediate planning problem. The Human Agentic Gap is the one forming behind it.

Every AI brand strategy investment made to date rests on a single assumption: that a human being is somewhere in the loop. Agentic commerce is removing that assumption. AI agents that run the full shopping flow autonomously — retrieval, comparison, criteria evaluation, transaction — without human review at each step are not a forecast. ChatGPT Shopping has integrated affiliate links. Perplexity has formalised its sponsored-answer programme. The major platforms are executing agentic commerce, not building towards it.

When the human is removed from the loop, three things disappear simultaneously:

•       Recovery opportunities. A brand displaced at turn two in a human conversation can re-enter if the consumer asks a follow-up question. In an agentic flow, displacement is permanent.

•       Diagnostic signal. The human's follow-up questions generate signal about where and why displacement occurred. Agentic flows generate none. The brand loses invisibly.

•       Time. A human journey across four turns takes minutes. An agent runs the same sequence in seconds. The window for Layer 3 activation narrows dramatically.

The GPT-5.5 update on June 25 sharpened this further. The model now generates 90% fewer follow-up questions — confirmed by OpenAI as intentional. The model is entering its forced-reasoning regime earlier. For low-consideration queries, the activation window is already compressing. For high-consideration categories, the mechanism driving reasoning is shifting from dialogue to inference, memory, and tool use.

The model is no longer gathering information through dialogue. It is gathering it before the conversation begins.

GPT-5.5 has been the default for every ChatGPT user since May 5. The compression is already the environment brands are operating in, not a future state to prepare for.

PART V · WHAT Q3 REQUIRES

Four things to do before September

1. Diagnose before you invest

The Linkage Gap and the Reasoning Gap require different remediation. A brand investing in Layer 3 activation to close a gap that is actually a structural fit problem will produce zero gap closure. The diagnostic question must be answered before budget is committed.

2. Separate your investment portfolio by axis

Possession, Layer 2, and Layer 3 investments are doing different jobs. Most current portfolios are over-weighted to possession and Layer 2. Most brands have no Layer 3 investment at all. A coherent portfolio is explicit about which investment is doing which job.

3. Build measurement infrastructure before agentic commerce arrives at scale

When an AI agent transacts on a consumer's behalf, current measurement infrastructure produces nothing actionable. The CDMO who commissions agent-level measurement now will have a proprietary advantage when competitors are scrambling to answer board questions about AI marketing ROI.

4. Watch model behaviour, not just platform rankings

The GPT-5.5 behaviour change is the signal that model architecture is moving faster than most AEO strategies account for. Brands tracking how model reasoning is shifting — not just which brands appear in which answers — will make better Layer 3 activation decisions than brands watching share of voice dashboards.

THE BOTTOM LINE

Cannes 2026 established that AI visibility is not optional. That consensus is correct and important. The next consensus — forming in Q3 — will establish that AI visibility is not sufficient.

The gap between being cited and being chosen is large, structural, and currently invisible to most brands. The gap between human-mediated AI journeys and autonomous agent journeys is opening faster than most planning frameworks account for.

The brands that build Layer 3 infrastructure now — before the category is crowded, before measurement standards are set, before the first mover advantage closes — will set the activation patterns their competitors then have to fight against.

The race to AI visibility is winding down. The race to AI activation is just beginning.

RESEARCH CITED IN THIS EDITION

Sheals, P. & de Rosen, T. (2026). Beyond Visibility: The Linkage Gap. AIVO Meridian White Paper v1.1. DOI: 10.5281/zenodo.14855818

de Rosen, T. (2026). The Human Agentic Gap. AIVO Journal, July 2026. DOI: 10.5281/zenodo.20920098

Jack, W. et al. (2026). Prominence-Stratified Failure Modes in Retrieval-Augmented Commercial Recommendation. arXiv:2605.27439

Conductor (2026). AI Brand Recommendation Study: Why Intent Type Predicts AI Output Consistency.

Havas (2026). The Science of Desire. havasscienceofdesire.com

Tim de Rosen CEO, AIVO, Inc.  · aivomeridian.com · tim@aivostandard.org

AIVO Meridian measures and closes the Linkage Gap for consumer-facing brands.