Edition One · November 2026

The Selection Index -
Which vendors AI recommends, one sector at a time.

Each sector measured the same way: when buyers ask AI to compare suppliers, who gets recommended, and which sources decide it.

Visibility tools count mentions. The Selection Index measures the decision: not whether a vendor turns up somewhere in the answer, but whether it makes the recommended set, and what the model read to get there. With the sector sets I keep between editions, it’s the evidence behind everything I do, for the vendors who want to be recommended and for the media and events businesses whose content the answers draw on.

First edition: fraud prevention and payment security. November 2026.

Free. You’ll get the edition the day it publishes, and The Selection Layer newsletter every fortnight after it.

Cover of the Fraud Prevention Selection Index, Edition One report

Page last updated: November 2026

What it measures

Measured from what the machines do, not from what people say

Each edition puts thirty questions a real buyer would ask to four AI engines every day for a fortnight, and to Claude by hand. The questions span the journey, from “who are the options” to “is this vendor reliable”, and the comparison and validation questions carry more weight, because that’s where shortlists get made. Every vendor in the sector is scored on whether it’s named in the answer, how it’s described, and whether it’s recommended or merely listed. The headline is the gap between named and recommended. Underneath sits the Source Layer: the domains the answers are built from, and how much of that is trade media. Owners of trade titles read that part first.

HOW IT’S RUN

  • Thirty questions in five classes: category discovery, context-rich buyer briefs, comparisons and alternatives, problem-led questions, and late-stage checks. Locked before the window opens, unchanged within it.

  • ChatGPT, Google’s AI Overviews, Gemini and Perplexity tracked daily, and Claude run by hand from dedicated accounts with memory off. The other four are captured automatically every day by a monitoring tool. That tool’s daily tracking doesn’t cover Claude yet, so rather than leave it out, I put the same questions to Claude myself on a fixed routine.

  • A fourteen-day window, so a score is a measure over time, not one answer on one afternoon. How much the answers moved is reported, not hidden.

  • A vendor counts as included only when it’s named in the body of the answer. A vendor named in the question itself doesn’t count on that question.

  • Positioning and Preference are coded by hand against a published rubric, and every code traces to a captured answer.

  • Where the numbers can’t separate two vendors, they’re shown as tied, not ranked.

  • One statement in every edition: these are baseline measurements from controlled sessions. Personalisation moves individual answers around that baseline, and no one can measure it.

Edition One

Fraud prevention and payment security, November 2026

The first edition measures the vendors merchants turn to for fraud prevention and payment security. It’s co-published with a leading payments title, which checked the questions for buyer realism before they were locked. It covers the vendors a merchant-side buyer would meet, from the category leaders to the challengers, scored the same way on every engine. The window ran from 12 to 25 October 2026.


For a vendor in the category, it answers whether you’re in the answer, whether you’re in the recommended set, and which sources put you there or left you out. For the media title covering the category, it answers how often the answers draw on trade media at all, and which kinds of source do the work instead.


No vendor paid to be in it, and none can. Edition One carries no sponsor. A vendor who questions a score gets the captured answers behind it, and a score changes only on a demonstrated coding error.


Register and the edition arrives the day it’s out. Subscribers to The Selection Layer newsletter get the full ranked table.

Edition One at a glance

November 2026

SECTOR

Fraud prevention and payment security

BUYER QUESTIONS

30, from category discovery to late-stage checks

AI ENGINES

5: ChatGPT, Google AI Overviews, Gemini, Perplexity and Claude

MEASUREMENT WINDOW

12 to 25 October 2026

CO-PUBLISHED WITH

A leading payments title

SPONSOR

None. No vendor pays to be in it

What you get from it

FOR MEDIA OWNERS

What the answers are built from

If you own a media title in the sector, the Index tells you how much of the shortlist conversation runs through trade media at all, which kinds of source the engines lean on instead, and whether your pages are being read without being named. That’s the Source Layer, and a board can’t get it anywhere else. The Source Snapshot is read from it, or from a one-off set I run for your sector where the Index doesn’t cover it yet: send me a title and you get a one-page extract and a thirty-minute readout, free.

FOR B2B VENDORS

Where you stand when the shortlist is built

If you market a vendor in an Index sector, the edition tells you whether you’re in the answer and whether you’re in the recommended set, against the competitors that matter, measured the same way. It doesn’t tell you why, or what to fix first. That’s the AI Shortlist Scan, which runs in any sector I can measure, is scored against the Index baseline where one exists, and comes with the sources, the beliefs behind each score and the five changes ranked by impact.

What’s inside

What’s free on this page, what newsletter subscribers get, and what only a Scan shows

Public, on this page

The sector story; the top three, shown as ties where the data can’t separate them; the headline chart, named against recommended; the domains that decide the category; the source mix by type of publisher; the patterns across vendors; how much the answers moved over the window; and the baseline statement.

For subscribers to The Selection Layer newsletter

The full ranked table, with each vendor’s composite score, Shortlist Rate and Inclusion by question type, and a one-paragraph read on each.

Only in a Scan, never in an edition

A vendor’s own Positioning and Preference detail, what the models believe about it and where each belief comes from, the appearance log behind every score, and anything that says what to fix.

The Index tells you you’re behind a competitor. The Scan tells you why, and what to fix first.

AI source mix by publisher type across six enterprise categories, UK, September to October 2026: trade media between 3% and 9% of sources cited

From the fortnightly tracking behind the Index. Edition One’s chart replaces this in November.

Next sectors

One sector a quarter, each with a media title that knows it

EDITION ONE · NOVEMBER 2026

Fraud prevention and payment security

With a leading payments title.

EDITION TWO · Q1 2027

One business technology category

Announced after Edition One, with a business technology title.

EDITION THREE ONWARDS

Nominated

Every edition needs a media title that knows the sector and buyers who use AI to compare suppliers in it. If that’s your category, nominate it. A nomination costs nothing and commits you to nothing. It tells me where the demand is.

Questions

Frequently asked questions

Is the Index free?

The public edition is. Register and it arrives on publication day. The full ranked table goes to subscribers of The Selection Layer newsletter, which is also free. No vendor paid to be measured.

Can a vendor pay to be included, or to improve its score?

No, and no. Every vendor a buyer would meet in the category is in, whether they like the result or not. Scores change only on a demonstrated coding error, with the captured answers to show it. If the answers change later, that’s movement, and it shows up in the next window.

How is this different from an AI visibility tool?

A visibility tool counts how often you’re mentioned. The Index measures whether you’re in the recommended set, codes how you’re described and why, and reports which sources the answer was built from. The gap between mentioned and recommended is the headline, and no mention count shows it.

Why only one sector a quarter?

Because each edition is thirty locked questions, five engines, a fourteen-day window and every recommendation coded by hand against a captured answer. That’s the only way the numbers survive a sceptical vendor reading their own line. One sector done properly beats ten done by dashboard.

Can I commission one for my sector?

If you own a media title in the sector, nominate the category and we’ll talk about co-publishing it. If you market a vendor, in this sector or any other I can measure, the AI Shortlist Scan measures your position now, in seven days, using the same method.

What happens to the data over time?

The questions for each sector are locked, so when a sector is measured again you’re reading movement, not a snapshot.

Register

Get the first edition the day it’s out

November 2026. Fraud prevention and payment security. Register and the edition lands in your inbox on publication day, with The Selection Layer newsletter every fortnight after it.

Own a media title in the sector? Send me the title and I’ll read you a one-page Source Snapshot before the edition is out.