Add pricing faq for PostHog AI credits

[annikaschmid]

Changes

• Adds a pricing page for AI credits pricing
• This is still WIP and I didn't consider navigation etc. I will work with @PostHog/content once I am happy with the contents

Add screenshots or screen recordings for visual / UI-focused changes.

Checklist

• Words are spelled using American English
• Titles are in sentence case
• Feature names are in sentence case too. It's "product analytics" not "Product Analytics" and so on.
• Use relative URLs for internal links
• If I moved a page, I added a redirect in vercel.json
• Remove this template if you're not going to fill it out!

Article checklist

• I've added (at least) 3-5 internal links to this new article
• I've added keywords for this page to the rank tracker in Ahrefs
• I've checked the preview build of the article
• The date on the article is today's date
• I've added this to the relevant "Tutorials and guides" docs page (if applicable)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Updated (UTC)
posthog	Ready	Preview	Nov 17, 2025 3:21pm

[patricio-posthog]

I think this section is confusing. As a customer I read everything but still don't understand the relationship between tokens and credits. If we can't be fully transparent here, I would remove any references to tokens.

[edwinyjlim]

I agree this brings up questions like "well, why aren't they just charging me by tokens then?"

That being said, I think this section is important to have. It just needs more detail and reasoning and ideally a conversion calculator or formula. I'm assuming we're charging a slight markup or something!

[edwinyjlim]

Also, what about input tokens vs output tokens? and tokens between different LLM models? This section probably needs an in-depth breakdown of a few different prompts and tasks.

[Twixes]

If we clarify the set markup, like @kappa90 points out on line 35, I think this becomes fairly clear and user-friendly:

AI credits are based on the underlying token costs. The choice of model is optimized automatically per PostHog AI feature, but in each case our markup is a constant 20% over the LLM provider price. This means that 1 PostHog AI credit = $0.8333 worth of raw inference.

Note that 25% markup would be a pleasant $0.8.

[annikaschmid]

Clarified in the relevant section below

[gewenyu99]

Since this is a draft, I won't have any detailed comments. This is more context for ya drive.

The convention is to have a "Cutting costs" section in most product sections. It's basically a pricing FAQ.

Generally we introduce:

• How the pricing works for this product
• Info about how to understand/estimate your bill
• Note about how we try to be cheap etc (just link to the blogs about our pricing)
• Ways to reduce your bill/cut cost (which is just like a spend wisely like this type of dea
• Lastly, FAQ catch all.

This is just context for you to consider <3

[edwinyjlim]

Exciting!

This is still WIP and I didn't consider navigation etc.

I'd focus just on the content! I have a PR coming soon that overhauls the PostHog AI docs and its layout, so we can merge your pricing content into that branch.

[edwinyjlim]

I agree this brings up questions like "well, why aren't they just charging me by tokens then?"

That being said, I think this section is important to have. It just needs more detail and reasoning and ideally a conversion calculator or formula. I'm assuming we're charging a slight markup or something!

[edwinyjlim]

Also, what about input tokens vs output tokens? and tokens between different LLM models? This section probably needs an in-depth breakdown of a few different prompts and tasks.

[edwinyjlim]

Exciting!

This is still WIP and I didn't consider navigation etc.

I'd focus just on the content! I have a PR coming soon that overhauls the PostHog AI docs and its layout, so we can merge your pricing content into that branch.

[kappa90]

This hides the markup calculation. We should be clear that we take the token cost given us by the provider, apply a markup and convert to credits.

[annikaschmid]

Clarified this

[kappa90]

I think we still don't mention the markup?

[Twixes]

This will be great to have, potentially to link to from the app too

[Twixes]

I don't think this is strictly true, i.e. if you getting help involves a query being created – that costs credits.

Because we have a (we could say generous) free tier, is it perhaps then simpler – for both the user an us – to charge for the root node always? Regardless of whether just the docs search tool was involved, insight creation, or no tools at all. Especially with the new agent orchestration based on mode-switching, being prototyped by @skoob13, the tools become "dumb", and it's even trickier to say "this is just getting help" vs. "this is doing some work for the user".

[tatoalo]

Yeah, completely agree @Twixes. We either are fully transparent on what we pass along to end user cost-wise or might as well remove this.
The current plan is to always charge for the root node anyway. Another (small) example is the title generator, we use a super small (i.e. fast) model to perform this simple operation, but we do still plan to pass that along to end user cost-wise. I would rather have a section explaining what we bill and what not, if not too technical and prone to being outdated (although we could work around that)

[annikaschmid]

The outcome of this discussion is that we've updated the calculation and will not charge for the root node when a user searches the docs.

[Twixes]

If we clarify the set markup, like @kappa90 points out on line 35, I think this becomes fairly clear and user-friendly:

AI credits are based on the underlying token costs. The choice of model is optimized automatically per PostHog AI feature, but in each case our markup is a constant 20% over the LLM provider price. This means that 1 PostHog AI credit = $0.8333 worth of raw inference.

Note that 25% markup would be a pleasant $0.8.

[pawel-cebula]

Maybe "inference costs" instead of "token costs"? Addresses some concerns raised in comments above + feels more accurate since IIRC some model providers have pricing components that are not purely token based

[pawel-cebula]

Will this real-time cost information include markup?

[annikaschmid]

yes 👍

[pawel-cebula]

add links

• https://app.posthog.com/organization/billing/overview
• https://app.posthog.com/organization/billing/usage)

[pawel-cebula]

Mention the default paid limit? Assuming we're going with it

[pawel-cebula]

Are these the standard billing alerts when you hit 80% and 100% of your free allowance (if on free plan) or billing limit (if on paid plan)? Or something custom for AI?

[annikaschmid]

Yes, standard. We don't have anything custom planned for now

[greptile-apps]

_{1 file reviewed, 8 comments}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

style: Add year to date for clarity (assuming 2024)

Prompt To Fix With AI

This is a comment left during a code review.
Path: contents/docs/posthog-ai/pricing-faq.mdx
Line: 22:22

Comment:
**style:** Add year to date for clarity (assuming 2024)

How can I resolve this? If you propose a fix, please make it concise.

[gewenyu99]

Nice some comments

This page also needs to be added to the side nav, but I'd suggest we just base this on top of #13614 and do the nav together there.

@edwinyjlim how's that sound

[gewenyu99]

Suggested change

	AI credits are consumed whenever PostHog performs intelligent work for you.
	You’ll spot these features by the ✨ icon or when using the in-app chat.
	AI credits are consumed whenever PostHog performs intelligent work for you. You’ll spot these features by the ✨ icon or when using the in-app chat.

@edwinyjlim maybe we rebase this on your other PR and link to the list of actions here (if we have a list)

[annikaschmid]

@gewenyu99 Thanks for your suggestions, I've committed them all. I left one comment on the calculation comment from you, feel free make the necessary change & commit directly, if it makes sense for you.

I would hand this now over to you and @edwinyjlim, so you can add it into the main AI docs. Thanks!

[edwinyjlim]

I would hand this now over to you and @edwinyjlim, so you can add it into the main AI docs. Thanks!

Sounds gyuuud. Consider this branch commandeered 🏴‍☠️ by me

[edwinyjlim]

incorporated into #13614