AI Credits

Every AI action in DateIQ — generating a reply, scoring a match, evaluating a profile — runs against an OpenRouter model and costs us money. Credits are how that cost is metered to your account.

The short version

1 credit = 1 AI call, with a few exceptions noted below.
Your balance lives in two buckets: plan credits (refilled monthly while your subscription is active) and pack credits (one-time top-ups that never expire).
Plan credits are spent first; packs cover the overflow.
Buttons that cost credits show a small sparkle icon and a number. No icon = free.
Run out and you'll get a clear "insufficient credits" message before the call goes out — nothing fails halfway.

The credits chip in the top-right of the extension header shows remaining / monthly grant. The colour goes amber when you're low and red when you're out. The same numbers — including pack balance and the next refill date — are shown on the Billing page in the dashboard.

What every action costs

Wingman (replies, drafts, openers)

Action	Cost	Notes
Suggest a reply	1	One model call per request. Each pattern variant costs the same.
Next action (autonomous)	1	Per generation, not per match. The wingman can be told how many to fan out.
Rewrite in a named style	1	E.g. Direct, Cold Read, Push-Pull.
Send-time lint of your draft	Free	Local heuristic — no model call.
Library openers (batch)	1	One request returns N openers in a single call.
Post-date opener	1	Uses a vision model to read context.

Coach (the chat panel)

Action	Cost	Notes
Coach reply	1 per turn	Tool calls inside a single turn (e.g. coach pulling up a match) don't add credits.

Profiles, photos, matches

Action	Cost	Notes
Score my profile (overall)	1	Cheap rubric model.
Evaluate profile (vision)	1	One model call regardless of how many photos.
Per-photo critique	1	One call returning all per-photo scores.
Score matches	1 per match	We batch internally for throughput; the price is per match scored, not per request.
Score one match	1	Used by the on-page ranker when re-checking a single profile.
Save a thumbs-up / thumbs-down on a match	Free	Pure feedback storage.

Wingman setup / wizard

Action	Cost	Notes
Extract preference axes from your input	1	Single AI call summarizing your free-form preferences.
Rate sample profiles in the wizard	Free	Only stores your ratings.
All other wizard steps (state, prefs, complete, abort)	Free	Plumbing, no model call.

Conversation context

Action	Cost	Notes
Conversation assessment (the "stage" tag)	1 on miss, free on cache hit	Runs the cheapest model and is cached per conversation hash. Most calls hit the cache.

Free always

Reading your matches, opening profiles, viewing past conversations — those are pure storage.
Pattern lists, coach tool registry, plan lookups, your account settings — all free.
Style flavour chips and post-date nudges are config, not generation.

Why some buttons cost more than others

A "1 credit per call" rule keeps the maths simple, but a few actions genuinely do more work in one shot. We follow two rules:

One model call = one credit. Whether the model is cheap (Haiku for stage assessment) or expensive (GPT-4o vision for photo critique), the user pays the same. Tier pricing absorbs the variance — that's the whole point of having tiers.
Batching is usually free, with one exception. When an endpoint bundles multiple items into a single call (e.g. all photos in one vision request, a batch of openers in one library request), the user pays for the call, not the items. Score Matches is the exception: even though we batch internally for throughput, the price is per match scored — one match analysed, one credit.

If a button doesn't show a credit chip, it's deterministic — no model involved.

Insufficient credits

When you're out:

The action returns a 402 and the UI surfaces an "out of credits" message with a link to top up.
We do not refund on AI errors. A pre-flight check rejects you cleanly when balance is empty, so the only way to "lose" a credit on a failure is if the model returns an error mid-stream after passing the gate. That's rare; if it happens to you and the result is empty or broken, email us and we'll add the credit back manually.
Your balance can briefly dip slightly negative on a single big call. We allow this on purpose — it's better than aborting an in-flight pipeline.

Plans, packs, and refills

Plan credits refill on the anniversary of your activation date. Unused plan credits expire at the refill — they don't roll over.
Pack credits are one-time purchases. They never expire and stack with whatever you already have.
Spending order: plan credits first, then packs. So if you buy a pack and don't use it this month, your plan credits still expire normally and the pack is untouched.
If your subscription lapses or you downgrade to free, plan credits go to zero immediately. Pack credits are unaffected.

The current pack catalogue is on the Billing page. Need a custom pack for very heavy use? Email us.

Unlimited tiers

If your plan is configured as Unlimited (e.g. the Ultra tier or staff accounts), every endpoint skips both the pre-flight check and the deduction. Your balance display shows "Unlimited" instead of a number. Usage is still logged for product analytics — it just doesn't bill.

Where the per-action cost is decided

The credit cost is set on the API side, not the client. The button's badge tells you what the server is going to charge — it's the same number the receipt shows in the response. If a future update changes a cost, the badge updates with it; we don't ship cost changes silently.

Help & FAQ