LawFather · Field Study
Back to the study
Study 04 · Full analysis

Five engines, five different answers. And they are not worth the same.

Research question: is there one AI ranking to win, or five separate ones, and does winning each one matter equally?

The hub establishes the headline: the engines barely agree, and they send wildly different amounts of traffic. This page goes further. Which engines agree with which, which are hardest to get named on, and the uncomfortable overlap that decides strategy: the two hardest engines to win are the two that send the most clients.

Method

How we measured whether two engines “agree”

For each question, every engine returns a list of firms. To measure agreement between two engines, we compared their lists for the same question and calculated the overlap: of all the distinct firms either engine named, what share did both name. A score of 100% would mean the two engines named exactly the same firms. A score of 0% would mean they shared no firm at all.

We did this for all ten engine pairs across every question, then averaged. We also counted, for each engine, how many of the 30 study firms it named at least once, which measures how hard that engine is to get named on at all.

One honest wrinkle. Engines differ in how many firms they list per answer. Most named eight to ten firms per answer. Gemini named far more, about 24, so it appears more “generous” and easier to show up on. We account for this below rather than letting it distort the strategy: naming a firm in a list of 24 is a weaker signal than naming it in a list of 8.

Part 1 · the disagreement, engine by engine

Even the two most-alike engines overlap only 16% of the time

The hub shows that 82% of named firms appear on only one engine. Here is why: no two engines are close to aligned. Ranked from the most-alike pair to the least, the overlap never rises above 16%, and ChatGPT sits at the bottom of nearly every pairing, meaning it is the engine that most goes its own way.

Agreement between each pair of engines  ·  share of named firms both engines share
Claude ↔ Perplexity
15.9%
Claude ↔ Gemini
15.5%
Claude ↔ Google AI
13.4%
Perplexity ↔ Gemini
11.6%
Gemini ↔ Google AI
9.8%
Perplexity ↔ Google AI
8.4%
ChatGPT ↔ Gemini
7.5%
ChatGPT ↔ Perplexity
6.4%
ChatGPT ↔ Claude
4.3%
ChatGPT ↔ Google AI
3.5%
Read this as: the top of the list, the engines that agree most, still overlap only about one time in six. The four lowest-agreement pairs all involve ChatGPT. The single most important engine for traffic is also the one whose recommendations look least like anyone else’s, so a strong showing elsewhere tells you almost nothing about how you are doing there.
Part 2 · the crossover that sets strategy

The engines easiest to win are the ones that send the fewest clients

Put two numbers side by side and the strategy writes itself: how hard each engine is to get named on, and how much traffic it actually sends. They run in opposite directions. Claude and Perplexity named the most firms and are the easiest to appear on, but send the least traffic. ChatGPT and Google are the strictest, and send by far the most.

Difficulty vs value, by engine  ·  firms named of 30 (our data) vs share of AI referral traffic (2026)
EngineFirms named /30Firms per answerReferral trafficWhat it means
ChatGPT128.4~75%Hardest and highest value. Win this first.
Google (Gemini + AI Overviews)1923.6~12%Second only to ChatGPT for traffic; names many firms per answer, so appearing means less.
Perplexity1810.2~7%Winnable, but a smaller slice of traffic.
Claude208.9~3%Easiest to appear on, least traffic sent.
Read this as: “firms named of 30” is how many of the study’s firms each engine named at least once, from our data, a proxy for how hard it is to get on that engine at all. The Google row’s difficulty figures (firms named of 30 and firms per answer) reflect the Gemini surface as tested. Referral-traffic shares are third-party 2026 estimates (SE Ranking).
The trap this exposes

A firm can celebrate a top spot on the engine that brings it almost no clients.

Because the engines disagree so completely, it is easy to look strong on Claude or Perplexity, where firms show up readily, and conclude you are winning AI, while being invisible on ChatGPT and Google, where three of every four AI-referred visitors actually come from. It is the old Bing problem in a new form: a number-one ranking is worth only as much as the traffic behind it. A single blended “AI score” hides this completely, which is why any honest measurement has to report each engine on its own.

The playbook

How to sequence an engine strategy

Measure each engine separately, always

Because agreement between engines tops out near 16%, a blended AI score is misleading. Track ChatGPT, Google AI, Gemini, Perplexity and Claude as five distinct scoreboards, and never assume a win on one implies anything about another.

Prioritize ChatGPT and Google first

They are the strictest to get named on and they carry the overwhelming majority of the traffic and reach. Effort spent becoming the answer here returns far more than the same effort on a lower-traffic engine.

Discount easy wins on low-traffic engines

Appearing on Claude or Perplexity is real, but weigh it against the traffic it actually sends. Do not let a strong showing on a minor engine create false confidence about your overall position.

Read Gemini’s generosity correctly

Gemini names roughly 24 firms per answer, so being one of them is a weaker signal than being one of ChatGPT’s eight. Treat a Gemini mention as a lower bar cleared, not as proof you have won.

LawFather

Want to know where you stand on each engine?

Our audit measures every firm engine by engine, so you see exactly where you appear on ChatGPT and Google, not just a single blended score. Run your firm’s audit and see the five scoreboards.

See your firm’s results