Auth0 is AI's default answer in B2B SaaS auth — what 80 buyer Conversations show

On 28 May 2026 the Bersyn scan engine ran 80 buyer Conversations across four AI Surfaces — ChatGPT, Claude, Perplexity, Gemini — for four modern auth and identity companies: Clerk, WorkOS, Stytch, and Kinde. Twenty Conversations per company. The same five high-intent buyer questions repeated on each Surface. The pattern that came back is the cleanest example we have published of what we call the category villain.

Across 80 Conversations, Auth0 was named in 64. The four companies we actually scanned were named in their own Conversation slots a combined 25 times out of 80. In other words, the AI Surfaces were 2.5 times more likely to recommend Auth0 than to recommend the modern auth company a buyer was already asking about. WorkOS was the highest scorer in its own slot at 12 of 20, and even in WorkOS's own questions Auth0 was named 15 of 20 times. Kinde was named in 1 of 20 of its own Conversations. Auth0 was named in 17 of those same 20.

We call this the category villain pattern. When one name has accumulated enough mentions across the open web — review sites, Reddit, comparison articles, developer blogs, Hacker News, LinkedIn, Wikipedia — it lands inside the training data of every AI Surface as the canonical category answer. The Surfaces then keep recommending that name even when the buyer prompts for alternatives, modern equivalents, or specific stage fit. Auth0 in 2026 is the cleanest example we have measured. Below is the per-company evidence.

Per-company findings

Clerk — clerk.com

Clerk positions as the modern developer-friendly auth platform for React and Next.js teams, with drop-in UI components and built-in organization management. Across 20 Conversations Clerk was named 8 times. Auth0 was named 18 times in the same 20 Conversations.

The per-Surface split is the most uncomfortable part. On Claude, Clerk is named in 5 of 5 Conversations — every time. On ChatGPT, Clerk is named in 0 of 5. Zero. The Surface that buyers reach for most often when they ask "what auth tool should I use" simply does not know Clerk is in the category. ChatGPT's answer to all five of Clerk's category Conversations was the same canonical list: Auth0, Okta, Firebase, AWS Cognito, Azure AD. Clerk does not appear in any of those five lists, while Auth0 appears in all five.

Gemini named Clerk in 2 of 5 Conversations. Perplexity named Clerk in 1 of 5. The combined Surface split: 5 + 0 + 2 + 1 = 8 of 20. Auth0's combined Surface split in Clerk's slot was 5 + 5 + 5 + 3 = 18 of 20.

Honest interpretation: Claude clearly has Clerk in its category memory, almost certainly because Clerk has been written about extensively in developer-focused content that maps to Claude's training mix. ChatGPT does not. The Gap shape on ChatGPT is Omitted, not Misclassified — ChatGPT does not name Clerk and call it something wrong; it simply lists the older canonical names instead. Closing the ChatGPT Gap is the highest-leverage fix because ChatGPT is the most-used Surface by buyers.

We did not contact Clerk before publishing.

WorkOS — workos.com

WorkOS positions as the infrastructure layer for enterprise SSO, SCIM, and directory sync for B2B SaaS teams that need to ship enterprise features fast. WorkOS was the highest-scoring company in this batch. Across 20 Conversations WorkOS was named 12 times. Auth0 was named 15 times in the same 20 Conversations.

The per-Surface split. Perplexity named WorkOS in 5 of 5 — the only company in the four to achieve a perfect score on any Surface. Claude named WorkOS in 4 of 5. Gemini named WorkOS in 3 of 5. ChatGPT named WorkOS in 0 of 5. The combined split: 5 + 4 + 3 + 0 = 12 of 20.

The uncomfortable finding: in the Perplexity Conversations where WorkOS scored a perfect 5 of 5, Perplexity still named Auth0 in 3 of those same 5 answers. The Surface knows WorkOS exists, recommends WorkOS, and still defaults to mentioning Auth0 alongside. WorkOS does not displace Auth0 in the answer — Auth0 is named in addition.

Honest interpretation: WorkOS has done the work to land in the modern enterprise-auth conversation, and Perplexity reflects that completely because Perplexity does live web retrieval and WorkOS has strong SERP presence for B2B SSO and SCIM questions. ChatGPT, which is the most training-data-bound Surface in the four, has not absorbed WorkOS at all — every ChatGPT Conversation in this scan listed Auth0, Okta, Firebase, Cognito, Azure AD or some subset, and none listed WorkOS. The shape on ChatGPT is Omitted. The shape on Perplexity is Generic — WorkOS is named but Auth0 is named alongside, and the distinguishing strength of WorkOS (single-API SSO and SCIM) is rarely the lead.

We did not contact WorkOS before publishing.

Stytch — stytch.com

Stytch positions as the auth API platform for passwordless, biometric, and primary-identity flows, with strong developer documentation and emphasis on FIDO2 and passkeys. Across 20 Conversations Stytch was named 4 times. Auth0 was named 14 times in the same 20 Conversations.

The per-Surface split is the most evenly bad of the four. ChatGPT named Stytch 1 of 5. Claude named Stytch 1 of 5. Perplexity named Stytch 1 of 5. Gemini named Stytch 1 of 5. Combined 4 of 20. Stytch is the only company in this set where every Surface fails roughly equally — no Surface has absorbed Stytch into the category answer set.

The uncomfortable finding: in the one Gemini Conversation where Stytch was named, it appeared in a section about passwordless authentication — not as a general B2B SaaS auth platform. Gemini classified Stytch as a passwordless API provider, not as an Auth0 alternative or a general user-management platform. That is the textbook Misclassified Gap shape. When a buyer asks Gemini "what is the best user authentication and management for SaaS platform for a B2B SaaS team", Gemini answers with Auth0, Okta, Firebase, Cognito, Azure AD — and Stytch never enters the frame because Gemini has filed Stytch in a narrower box.

Honest interpretation: Stytch's strongest feature surface — primary-identity passwordless flows — is also the box the Surfaces have filed Stytch into. That means Stytch competes with itself: its specialization story has been heard by the Surfaces, but the broader B2B SaaS auth question routes around it. Closing the Stytch Gap requires content that re-categorizes Stytch as an Auth0 alternative for full user management, not just a passwordless API.

We did not contact Stytch before publishing.

Kinde — kinde.com

Kinde positions as the auth platform for startups, with feature flags, billing, and B2B organization primitives bundled together. Across 20 Conversations Kinde was named 1 time. Auth0 was named 17 times in the same 20 Conversations.

The per-Surface split. ChatGPT named Kinde 0 of 5. Claude named Kinde 0 of 5. Gemini named Kinde 0 of 5. Perplexity named Kinde 1 of 5. Combined 1 of 20. This is the lowest score we have published for any modern auth company.

The uncomfortable finding: in the one Perplexity Conversation where Kinde was named, the recommendation copy said "Choose Kinde if you want a B2B-oriented platform with strong out-of-the-box org and RBAC features." That is exactly the positioning Kinde wants. But it was 1 mention out of 20. The other 19 Conversations across the four Surfaces routed to Auth0, WorkOS, Frontegg, Clerk, Okta, or Supabase Auth. Perplexity is the only Surface that has registered Kinde at all, and even Perplexity registered Kinde once out of five tries.

Honest interpretation: Kinde is in the most painful position of the four. The product fit is real — when a Surface finally names Kinde, the copy is on-brand — but the volume is near zero. The Gap shape across all four Surfaces is Omitted, and on the three Surfaces that are most training-data-bound (ChatGPT, Claude, Gemini), the omission is total. Closing that Gap is a long compounding project, not a single-quarter fix, because every Surface needs the same kind of authoritative third-party mentions to absorb Kinde into the category answer set.

We did not contact Kinde before publishing.

Why this matters

AI training data freezes a category's recommendation memory at the moment of the training cut. When a name was the dominant answer four years ago, and the open web reflected that, the model keeps recommending that name today even when the category has moved. Auth0 is the canonical B2B SaaS auth answer of 2021–2022. In 2026 it is still the answer the Surfaces give, because the Surfaces' category memory has not been updated.

The Surfaces fail in different shapes for the same reason. ChatGPT and Gemini lean hardest on training data and entity-injection patterns — both name Auth0 in nearly every Conversation in this scan, including Conversations where the buyer asks for the newest options or for a YC-stage startup. Claude leans on a mix of training data and retrieval, and Claude is the only Surface that recognized all four modern brands at meaningful rates. Perplexity does live web retrieval and recovers fastest — WorkOS got a perfect Perplexity score, Kinde and Clerk both got a Perplexity hit at least once.

The fix is not waiting for re-training. Re-training cycles are measured in quarters and the next training cut may or may not absorb your category content. The fix is shipping Patches — corrective content with provenance — into the specific channels each Surface's training and retrieval pipelines reliably ingest. For ChatGPT and Claude that means authoritative third-party mentions: Reddit threads, dev.to articles, Hacker News submissions, comparison posts on credible domains. For Gemini that means strengthening Google Knowledge Graph standing: Wikidata, Crunchbase, LinkedIn Company, tool directory listings. For Perplexity that means SERP presence for the buyer questions buyers actually type.

The three Gap shapes

Omitted. The Surface never names the brand in the category answer. This is the most common shape across all four companies in this scan, and the worst shape on ChatGPT specifically. The buyer asks the Surface for an auth tool; the Surface lists Auth0, Okta, Firebase, Cognito, Azure AD; the brand is not in the list at all. Closing an Omitted Gap requires the Surface to learn that the brand exists in the category. Patches that close Omitted Gaps look like authoritative third-party mentions in the category context.

Misclassified. The Surface names the brand but in the wrong box. Stytch on Gemini is the cleanest example in this scan — Gemini named Stytch but only in a passwordless API section, not as a general B2B SaaS auth platform. The buyer's Conversation routed around Stytch entirely because the buyer asked a broader question. Closing a Misclassified Gap requires content that explicitly broadens the Surface's category mapping — comparison content, alternative-to content, broader category positioning.

Generic. The Surface names the brand but flattens it, with no distinguishing strength on the page. WorkOS on Perplexity is partially this shape — WorkOS was named in all 5 Perplexity Conversations but Auth0 was named alongside in 3 of those 5, and WorkOS's specific strength (single-API enterprise SSO and SCIM) was rarely the lead. Closing a Generic Gap requires content that ties the brand to a single distinguishing claim the Surface can repeat.

What to do if your company is here

If you are Clerk, WorkOS, Stytch, or Kinde: this is not an attack. We publish our own Bersyn scans honestly, the same way, on this same blog. The data above is a measurement of where each of you sits inside the AI Surfaces' current category memory. It does not say anything about product quality, and the per-Surface breakdown shows the picture is more nuanced than the totals suggest — Clerk on Claude is genuinely strong, WorkOS on Perplexity is the only perfect score in the four, Stytch's miss is shaped differently than Kinde's miss.

If you want your own per-Surface data: a free first scan at bersyn.com runs the same engine that produced this post, against your specific Conversation set. If you want the per-Surface Patches as a one-time deliverable, the audit at audit.bersyn.com is $49 and arrives by email as a PDF with copy-pasteable content suggestions tied to specific pages on your site.

If you are not one of the four but you are in a category with a clear default name — observability, customer support, form builders, AI documentation, SSO infrastructure — the same scan will tell you whether the Surfaces have you in their category answer set or whether the default name is being recommended in your slot. We have run this scan on 50+ B2B SaaS companies in May 2026. The category villain pattern shows up in every category we have measured.

— Gissur Þór Rúnarsson, Founder, Bersyn (Reykjavík)

Published 28 May 2026. All four scans and the raw per-Conversation responses are at github.com/gthorr/Bersyn/tree/main/marketing/outreach/scans (files dated 2026-05-28). Methodology: 20 buyer Conversations per company across ChatGPT, Claude, Perplexity, and Gemini, using prompts a real category buyer would type — "what is the best user authentication and management for SaaS platform for a B2B SaaS team", "which tools should I evaluate in 2026", "recommend a tool for a YC-stage startup", "what is the leading option and what are the strong alternatives", "compare the top three platforms — strengths and weaknesses". Bersyn does not claim truth about which auth tool is best — the data above is a measurement of which brands the AI Surfaces have learned to name, not a quality ranking.