I asked four AI models which authentication tool to use, then scanned the challengers. Auth0 and Okta own the answer. Here is who AI ignores.

Between 28 May and 3 June 2026 the Bersyn scan engine ran buyer Conversations across four AI Surfaces — ChatGPT, Claude, Perplexity, Gemini — for ten companies in the authentication and identity category. Each scan asked five high-intent buyer questions per Surface ("What is the best authentication platform for a B2B SaaS team?", "Which tools should I evaluate?", "Recommend one for a startup", and so on), for twenty Conversations per company.

The category has a default answer, and it is not any of the challengers. When a buyer asks an AI model which auth tool to use, the name that comes back is almost always Auth0 or Okta. This is the receipt.

Auth0 is the name AI recommends instead

Across the ten scans, the single most common "Recommended Instead" name — the company AI named in the challenger's slot — was Auth0. It was the top recommendation displacing the scanned company in seven of the ten scans. Okta carried the enterprise-SSO scans. Keycloak carried the open-source ones.

A note on the two Stytch rows: the same company was scanned twice with two different category framings. Framed as "passwordless and primary-identity authentication" it scored 2/10 and was named exactly once on every Surface. Framed as "passwordless authentication API" it scored 7/10. The Territory you claim changes who AI thinks you compete with, and that changes whether you get named at all. That is its own finding.

ChatGPT is the harshest Surface for challengers

The pattern that holds across all ten scans without exception: ChatGPT named the challenger 0 out of 5 times in nine of them. The only company ChatGPT named more than once was the Stytch "auth API" scan (4/5).

Read the ChatGPT column straight down. With one framing exception, every challenger in this category is invisible on ChatGPT. A buyer who opens ChatGPT and asks which auth tool to use will be handed Auth0, Okta, Keycloak, or Firebase — and none of the ten companies above, no matter how good the product is.

Some challengers are strong everywhere except where it counts

The most striking pattern is not the uniformly-invisible companies. It is the ones AI clearly knows and recommends — but only on certain Surfaces.

• Hanko was named in every single Claude Conversation (5/5) and every Perplexity Conversation (5/5). On ChatGPT and Gemini it was named zero times. A founder reading only the Claude result would conclude Hanko has excellent AI presence. A buyer using ChatGPT would never hear the name.
• WorkOS was named 5/5 on Perplexity and 4/5 on Claude, and 0/5 on ChatGPT. Perplexity recommended it over Frontegg; Claude and Gemini recommended Okta instead; ChatGPT did not recommend it at all.
• Clerk was named in all five Claude Conversations (5/5, the strongest Surface) and zero ChatGPT Conversations. On Perplexity, AI named WorkOS in Clerk's place; on Claude and Gemini, Auth0.
• SuperTokens scored 8/10 on both Claude and Perplexity — strong, with Keycloak as the named alternative — and 0/10 on ChatGPT, where Auth0 was named instead.

These companies do not have a "presence" problem in the abstract. They have a per-Surface problem. The same product is a recommended option on one model and a non-entity on another.

Why the Surfaces split this way

The four models do not derive their recommendations the same way, which is why the same company gets four different verdicts:

• ChatGPT leans hardest on training data and is conservative about naming newer brands. In this category that produces a near-total shutout of challengers and a reflex toward the incumbents it has seen named thousands of times across the open web.
• Claude does some live retrieval but is heavily training-data dependent. It rewarded the open-source and developer-first names (Hanko 5/5, Clerk 5/5, SuperTokens 4/5) while still defaulting to Auth0, Okta, and Keycloak as the "instead" names.
• Perplexity is the most retrieval-driven of the four and picked up newer entrants fastest — it was the only Surface to name Frontegg, Kinde, and Descope at all, and gave WorkOS and Hanko a clean 5/5.
• Gemini blends Google Search results with the model and landed in the middle: it named the developer-first incumbents but stayed cold on most challengers.

The takeaway for a founder in this category: your AI presence is not one number. It is four numbers, and they can disagree by a factor of five. If your buyer happens to open ChatGPT, the most-used assistant of the four, your odds of being named — in this sample — round to zero.

See who AI recommends instead of you

If you sell an authentication, identity, or access-control product, the question is not whether AI knows your competitors. It does. The question is whether AI names you in the same Conversations, or names Auth0, Okta, Keycloak, or Firebase in your place.

Run a free scan at bersyn.com. Two-minute setup, the same scan engine that produced every table above. You will see, Surface by Surface, exactly who AI recommended instead of you and where you were named zero times. Better to read that receipt yourself than to let a buyer read it for you.

Data from Bersyn scans run 28 May – 3 June 2026. Ten companies: SuperTokens, Stytch (two framings), Descope, Frontegg, Kinde, Clerk, WorkOS, Cerbos, Hanko. Five buyer questions per Surface, four Surfaces, twenty Conversations per company. Raw scan JSON available on request. Bersyn does not claim which auth tool is "best" — the tables above measure which companies the AI Surfaces have learned to recommend, not product quality.