The Annual · Edition I

The State of AI Discovery.

An annual reckoning of how AI systems answer commercial questions — what they cite, who they surface, where they err, and how the discoverability of businesses inside them is changing year over year.

State of AI Discovery
2026
The first annual edition. Published February 2026.
Founding Edition
What it is

The reference document for the field.

Each February, we publish the only longitudinal record of how AI systems behave as commercial discovery infrastructure. The Annual is read by founders, marketers, agencies, regulators, journalists, and academics. It is the record, not the commentary.

Part I

The Year in AI Discovery

The headline movements — model behaviour shifts, citation pattern changes, regulatory developments, the year's notable absences and arrivals.

Part II

The Index, in Aggregate

What the AI Discovery Score looks like across industries, geographies, and business sizes. Where the gap is widest. Where it is closing.

Part III

The Eight Signals, in Detail

Annual analysis of each Discovery Signal — Mention Frequency, Authority, Citation Quality, Entity Resolution, Schema, Authority Graph Distance, Behavioural Reinforcement, Recency Stability.

Part IV

The Verticals

Vertical-by-vertical readouts for legal, accounting, dental, real estate, financial advice, allied health, construction, and the other indexed industries.

Part V

Methodology

The full methodology document for the year. The Source Register summary. Changes from the prior edition. Editorial governance and conflicts disclosed.

Part VI

Outlook

Where the institution expects the field to move in the coming year. Hypotheses to test. Questions for the next Annual.

2026 Findings

Six findings the field could not ignore.

The 2026 edition is shaped by the first full year of post-search AI discovery. These are the findings most cited by the press, the regulators, and the operating teams who have read it.

Finding 01

Citation concentration is rising.

Across the indexed industries, the top fifty cited sources now carry a meaningfully larger share of AI answers than they did in late 2024. Smaller, independent sources are being deprioritised at scale.

Finding 02

Entity confusion is the largest single cause of misattribution.

More than four in ten observed AI errors against indexed subjects trace to entity resolution failures — the system could not determine which legal entity the answer described.

Finding 03

The Index gap widened in three industries, narrowed in two.

Cosmetic clinics, real estate, and construction widened. Legal and accounting narrowed. The Annual analyses why, and what is likely to change in 2027.

Finding 04

Behavioural reinforcement is the most volatile signal.

Recency-weighted review and citation behaviour now changes scores faster than any structural signal. Operating teams are advised to instrument it.

Finding 05

Authority Graph Distance is the most stable predictor.

Where Mention Frequency rises and falls with content cycles, distance to Tier I authority sources predicts AI placement most reliably across all eight indexed verticals.

Finding 06

The institutions absent from AI answers are not the ones you would expect.

The institutions most absent from AI answers are not the smallest or the youngest — they are the ones with the strongest offline reputations and the weakest authority architectures.

How it is made

One year of work, four phases of review.

The Annual is not a quarterly trend report. It is the single canonical document of the year, prepared with the editorial discipline of a financial publication.

Phase 01 · March–August

Drafting

The institution drafts each part across the operational year, capturing what the Index, the engagements, and the Methodology Council have observed.

Phase 02 · September

Drafting closes

Drafting closes 1 September. From this date the document is in editorial review only; no new findings are added to the body of the edition.

Phase 03 · October

Editorial committee

The editorial committee — the Methodology Council and the Head of Practice plus one external observer — reviews the full draft against editorial standards and ratifies the findings.

Phase 04 · November–February

Final ratification, publication

Final ratification 1 November. Publication first business day of February. The Annual becomes the canonical reference for the field for the year.

Editions

Two editions. One canonical record.

The institution publishes the Annual in two editions, the same document across both, distinguished by access and physical form.

Public Edition

Distributed to founders, marketing leaders, agencies, journalists, and academics. PDF and web edition, fully indexable. Methodology and findings in full.

Available at no charge to qualifying recipients. Delivered February each year.

No chargeto qualifying recipients

Institutional Edition

Hardback printed edition, individually numbered, archive-grade paper. Includes the year's Source Register in full, methodology supplements, and the verbatim minutes of the editorial committee.

Distributed to subscribing institutions, regulators, and academic libraries. Limited print run.

$3,500/year subscription
Publication record

The first edition is the founding edition.

Each edition is a numbered volume in the institution's permanent record. The publication record is itself part of the methodology — the institution holds itself to a public schedule.

2026
State of AI Discovery 2026 — Founding EditionPublished 2 February 2026. Eight indexed verticals, eight Discovery Signals, six findings of record, full methodology and Source Register.
Edition I
2027
State of AI Discovery 2027To be published February 2027. Drafting opens March 2026.
In progress
2028
State of AI Discovery 2028To be published February 2028.
Scheduled
Request the Annual

Read the record of the year.

The 2026 Founding Edition is available now to qualifying recipients. The Institutional Edition ships in archive-grade hardback. Either way, the Annual is the document the field reads first.

Request the Annual