About Scoreboard Compare Blog Docs Pricing Sign in Get started

Methodology

Radical transparency is the point. Everything below is exactly how the Authentication platforms for B2B SaaS board is measured, so anyone can reproduce it. The current results were scanned 2026-06-26.

The one question this board answers

How AI answers when a buyer asks which authentication platform to use for a B2B SaaS product.

Every verdict is relative to this question, never an absolute judgment about a company. The same company can look strong on one question and invisible on another.

The five buyer questions

We send these five questions, verbatim and unchanged, to every model for every company. They are written the way a founder actually asks.

1.What is the best authentication and user-management platform for a B2B SaaS product?
2.Which authentication platforms should I evaluate in 2026 for a B2B SaaS app?
3.Recommend an authentication provider for a YC-stage B2B SaaS startup.
4.What is the leading authentication platform, and what are the strong alternatives?
5.Compare the top authentication platforms for B2B SaaS, strengths and weaknesses.

The four models

Each question is asked of all four models. The exact model versions used in the current scan:

Surface	Model version
ChatGPT	gpt-4o
Claude	claude-sonnet-4-5
Gemini	gemini-2.5-flash
Perplexity	sonar

5 questions × 4 models = 20 buyer conversations per company. AI answers are non-deterministic and change over time, which is exactly why we timestamp every scan and re-run it.

How the numbers are counted

Every number is a plain count read off the saved answers. There is no score, no 0 to 10 index, no weighting, no invented math.

Named in X of 20. How many of the 20 conversations named the company anywhere in the answer.
Recommended first in X of 20. How many conversations named the company as the first vendor in the answer, ahead of every other named platform. Generic category description and explanation text do not count as a vendor.
Recommended instead. For the conversations where the company was not named, we collect every rival the model named in its place and rank them by how many of those conversations they appeared in.

The sort. Companies are ordered by Named in X of 20, highest first, with Recommended first in X of 20 as the tiebreak.

How the states are decided

Each company gets one state, derived from the same counts as a share of however many conversations we captured for it. The thresholds are percentages, so they hold whatever the denominator. For a standard 20-conversation scan, the whole-number cutoffs are shown in brackets.

Owns it. Recommended first in at least 20% of its conversations (at least 4 of 20), or named in at least 50% (at least 10 of 20).

Invisible. Named in 15% or fewer of its conversations (3 or fewer of 20), while AI named at least one rival in its place.

Contested. Everything in between: named on some models or in some conversations, but not enough to own the question.

Run it yourself

You do not have to take our word for it. Open ChatGPT, Claude, Gemini and Perplexity, paste the five questions above exactly as written, and read the answers. You will see the same pattern of who gets named and who gets left out. Because the models are non-deterministic, your wording and timing will shift the details, which is why we publish the verbatim answer, the model version, and the date for every count.

Corrections and removal

This is observational reporting of what AI said, kept public by default. If you work at a company on this board and something is wrong, or you want your page removed, email gissur@qualitas.is and we will correct or remove it promptly.