Back to the blog
GEOResearchstrategytactics

AI has picked your Auth provider and you didn't notice

Tom WalshTom Walsh
5 min read

A developer opens ChatGPT and types "best authentication tool for a Next.js app." Three seconds later they get a ranked list of names. They click one. They never see the other twelve.

You weren't in the meeting. You weren't in the search results either. The AI made the shortlist, and you either made the cut or you didn't.

So I pulled the data. We ran 10 buying-intent prompts across 4 AI surfaces, 40 answers total, and tracked who got named and, more important, what the AI read before it answered. The results should change where you spend your content budget this quarter.

The shortlist is brutal, and it's already half-decided

Clerk took 25% of first-place mentions. Next.js/Auth.js and Auth0 each took 12.5%. That's three names owning half of every "who's first" slot in the category.

The median answer named about 12 tools. Sounds roomy, until you remember 38 of 40 answers came back as a ranked list, and nobody reads one down to position 9. Past roughly position 3, you're a footnote the buyer scrolls past. You got named and you still lost.

Here's the part that should keep you up at night, and the part you can actually do something about. 17 different tools landed in first place at least once. The top 3 own half the mentions, but 14 other tools are splitting the rest right now. The order isn't locked. That's the opening.

What the AI actually read before it answered

This is the finding I'd staple to your content lead's monitor.

We counted every source the AI cited across all 40 answers. Here's the breakdown:

SourceCitationsShare
Vendor blog posts12244.5%
"Other" web7928.8%
Reddit3512.8%
YouTube165.8%
GitHub82.9%
Vendor docs62.2%

Vendor blog posts: 122 citations. Vendor docs: 6. Blogs out-cited docs 20 to 1.

31 of the 40 answers cited a vendor blog. Only 6 cited anyone's docs. Documentation finished dead last out of every source type we tracked.

Some of you are thinking: but our docs are incredible, we spent a year on them. I believe you. Your docs are great for the developer who already picked you and is wiring up a login flow at 11pm. They do almost nothing to get you picked in the first place. When a buyer asks who's best, the AI reaches for the comparison post. Your API reference never enters the picture.

Why Clerk won, in Clerk's own words

We checked the actual URLs.

In the answers where Clerk landed first, the AI cited Clerk's comparison articles, not its docs. Pages like clerk.com/articles/...what-authentication-solutions-work-well and ...top-tools-and-recommendations-for-2025. Clerk wrote the "here are the best auth tools" article, and the AI quoted Clerk's article back to the buyer with Clerk sitting at the top of it.

The same fingerprint shows up on Kinde, WorkOS, Descope, and Stytch. Every one of them showed up through "top 10 auth tools" and "best auth for X" blog content. The vendors writing the roundup are the vendors landing in the roundup.

The companies winning the AI recommendation are the ones who sat down and wrote the recommendation themselves.

The surface makes it worse

The bias gets worse on the surfaces where your buyers actually are.

Citation depth per answer: Google AI Overview pulls about 9.7 sources, Claude search about 8.8, ChatGPT search about 5.0, Google AI Mode about 3.9. So a developer using Google's AI Mode gets an answer built on roughly 4 sources, and every one of those slots still skews blog-first. If you're not in the handful of pages that surface reads, the answer never mentions you exist.

What I'd do Monday morning

Stop treating "write the comparison post" like a dirty job you hand to the intern. It's the single highest-impact piece you can ship right now. Here's the order I'd run it in.

Write the roundup you want to win. Publish your own "best authentication tools for [specific use case]" article. Pick the use cases your ICP actually searches: best auth for Next.js, best auth for B2B SaaS, best auth for React Native. Yes, you're in it. So is the honest competition. The AI rewards the page that compares fairly.

Make it a real comparison or skip it. A roundup that lists you at #1 and everyone else as garbage gets ignored, by the AI and the reader. Name competitors. Give them fair credit. Be specific about who you're not for. A buyer asking "best auth for a 3-person startup" should find a page that honestly says when Auth0's pricing stops making sense and when yours starts. That honesty is exactly what gets cited.

Use the words your buyers type. "Top tools and recommendations for 2025." "What authentication solution works well for X." Those phrasings aren't an accident in the winning URLs. Match the question, get pulled into the answer.

Don't kill your docs. Reassign them. Docs close the developer who already chose you. Comparison content gets you chosen. You need both jobs done, by two different pages, aimed at two different moments.

The category is consolidating fast. The top 3 already own half the first-place slots, and every quarter you wait, those names collect more citations and more reasons for the next AI answer to repeat them. The window where 14 tools are still fighting over the back half of the list is open right now. It will not stay open.

Go write the roundup. 🎯