What AI features actually cost — the credit is where the price hides
Every tool you already pay for has grown an AI feature, and almost none of them are priced like the rest of the product. We went back through our own pricing research on 300 tools and pulled out every place where the AI is billed separately. 22 of them had something specific enough to quote. Those quotes are below, word for word from our dossiers.
One subscription, several meters
When software was software, the price was per seat and you could multiply. AI features broke that, because the vendor's own cost is per generation. So the bill splits into meters that run at different speeds, and the one on the pricing page is usually the slowest.
Our illustration. The four patterns below are what we actually found in our research — not a framework we invented and then went looking for.
1. Credits, where a retake counts as a use
This is the big one, and the trap inside it is not the price per credit — it's that nobody produces anything on the first attempt. Our ElevenLabs dossier says it as plainly as we can: credits are the meter and retakes are the tax. The same shape shows up in agent tools, where one instruction is not one credit: Lindy.ai burns 5–10+ credits on a research-heavy run, and Emergent 10–20+ on a complex change. If you budget by finished outputs, you will be wrong by whatever your redo rate is.
2. Caps per tier, which decide your ceiling
Here the AI is included, but rationed, and the ration is often what actually separates the tiers. Descript runs two meters underneath a per-editor price: media hours and monthly AI credits. Gamma gives 400 one-time credits free, then 1,000 a month on Plus and 4,000 on Pro. Miro's free plan carries 10 AI credits a month. None of that is hidden — but it means the tier you need is set by your usage, not your headcount.
3. AI as an add-on, priced apart from the plan
Aircall notes that AI features increasingly live in add-ons or higher tiers. Later lists AI credits among its add-ons. Wati meters automation triggers and AI credits separately, with Meta's per-conversation fees landing on top. The pattern to watch: if the AI is the reason you're buying, check whether it is inside the number you're comparing.
4. Nothing published at all
The most uncomfortable one. Pingtree is the clearest example we found: its entire homepage is built around Copilot, its prompt-driven AI agent, while the pricing page is written purely in infrastructure units — leads, requests, clicks, storage, API calls — and says nothing about AI usage limits or which tier gates which Copilot capability. That is not a criticism of the product. It is a warning that you cannot budget it from the website.
What our dossiers say, tool by tool
Each quote below is lifted verbatim from our pricing research for that tool. We publish these because the number on the pricing page rarely survives contact with real usage — and because a caveat is worth more than a headline price.
Aircall“'Unlimited calling' covers listed regions — check yours — and AI features increasingly live in add-ons or higher tiers.”
Atria“It's credit-metered — AI credits and brand-tracking slots cap what you can generate and monitor, so a heavy user burns through Core's allowance and steps up to Plus.”
Bread & Butter“Two meters to watch: it prices per WEBSITE (multi-site setups multiply the $49) and the AI features run on compute-credit blocks that stack as usage grows.”
Descript“It's per-editor pricing with two meters running underneath: media hours AND monthly AI credits.”
ElevenLabs“Credits are the meter and retakes are the tax — regenerate a paragraph three times and you pay three times, so real usage lands well above the calculator estimate.”
Emergent“The real cost is credit consumption, not the flat sticker — a genuinely complex app burns 10–20+ credits per big change, so serious builders move up tiers faster than the price list implies.”
Fireflies.ai“Paid tiers cap AI credits, and if you mainly want real-time in-Zoom captioning and simple English notes, Otter is the simpler pick.”
Freshchat“The free tier is genuinely useful for small teams, but AI features and higher-tier routing sit behind paid plans and add-ons — and being part of the wider Freshworks suite means you'll be nudged toward the full platform.”
Freshservice“Price the tier that contains your must-haves plus realistic AI usage, not the starter row.”
Gamma“Free (400 one-time AI credits, Gamma branding); Plus about $12/mo (~$9 annual, 1,000 credits/mo, no badge), Pro about $25/mo (~$18 annual, 4,000 credits), plus Team and Business tiers (2026).”
InVideo“The honest catch is the credit model — AI minutes are metered and don't roll over, so heavy users burn an allowance faster than the flat monthly price implies, and the free tier's watermark makes it evaluation-only.”
Lindy.ai“The credit model is the catch — complex, multi-step or research-heavy agents burn 5-10+ credits per run, so real usage can outpace your plan fast.”
ManyChat“The AI add-on is closer to keyword-matching than real conversational AI, billing scales with contacts (a viral Reel can spike your bill), and support is email-only and slow.”
Miro“(2026) A free plan (3 editable boards workspace-wide, unlimited members, 10 AI credits/mo); Starter about $8/member/mo billed annually ($10 monthly), Business $20/member/mo annual ($25 monthly), Enterprise custom from 30 seats.”
Ngram“And AI video is fast but templated by default: turning docs and screenshots into a decent clip is genuinely useful, but it needs your brand direction to not look generic, and it won't replace bespoke studio production.”
Pingtree“Copilot, the prompt-driven AI agent, is what the entire homepage is built around, yet the pricing page is written purely in infrastructure units — leads, requests, clicks, storage, API calls — and says nothing about AI usage limits or which tier gates which Copilot capability.”
Rank Math“The free version is so generous that many sites never need to upgrade — PRO mainly earns its keep once you want built-in rank tracking, Content AI credits or to manage many sites.”
SEO Space“Read the ladder carefully before you jump a tier: the 3-site Pro plan carries the same published limits as the 1-site Starter — 100 pages, 10 AI credits, 25 tracked keywords, monthly monitoring — so the extra $31 a month buys site slots, not capacity.”
Softr“Generous free plan (one published app, 10 users, AI credits); Basic about $49/mo annual (roughly $59 monthly), Professional around $139/mo, Business about $269/mo (2026).”
Team Pulse“Paid (billed annually): Starter $29/mo ($348/yr) — 1 member, 5 projects, NO pipelines; Growth $79/mo ($948/yr) — 3 members, 10 projects, 1 pipeline, AI; Pro $159/mo ($1,908/yr) — 10 members, 20 projects, 2 pipelines; top tier $299/mo ($3,588/yr) — 40 projects, 1,500 AI credits.”
Wati“Automation triggers, AI credits, the Shopify link and every additional WhatsApp number are all separately metered, and Meta's per-conversation fees land on top of the lot.”
Quotes are taken from our own dossiers and change when the research does. Prices move constantly: verify the current numbers on each vendor's pricing page before you commit. The full research for each tool sits in its dossier, and the cost-trap flags across all 300 tools are in the Atlas.
How to budget it without guessing
Four questions, in order. They take about ten minutes per tool and they are the ones that would have saved us money.
- What is one unit of my work worth in credits? One finished deck, one minute of usable video, one enriched contact. If the vendor publishes the conversion, this is arithmetic. If not, ask them — and note that having to ask is itself information.
- What is my redo rate? Multiply by it. Two or three is normal for anything creative.
- Does the allowance renew or is it a one-off? "400 credits" and "400 credits a month" are different products.
- Is the AI in the tier I'm comparing, or one above? This is the single most common reason a quote comes in higher than expected.
And one thing not to do: don't pick the tool with the most generous-sounding credit allowance. Allowances are denominated in each vendor's own unit, so they aren't comparable until you convert them. A big number in a cheap currency is still a small budget.
Frequently asked questions
Why do AI tools charge in credits instead of a flat fee?
Because the underlying cost is per-generation, not per-user. A model call costs the vendor real money every time it runs, so flat pricing would lose money on heavy users. Credits pass that variability on to you. The honest version of this would be a clear conversion rate; the common version is a number with no stated exchange rate.
What is the most expensive mistake when budgeting for AI features?
Budgeting the sticker price and forgetting the retake. Nobody generates once. As our ElevenLabs dossier puts it, retakes are the tax — regenerate a paragraph three times and you pay three times. Estimate your real number of attempts, not your number of finished outputs.
How do I compare two tools when both price in credits?
Convert both to your own unit of work: cost per finished deck, per minute of video, per enriched contact. That means finding out what one unit consumes, which vendors often bury. Where a vendor publishes it — Ngram states roughly 50 credits per export and about 10 per second of AI video — comparison is easy. Where they don't, treat the silence as a risk, not a detail.
Are free AI tiers worth anything?
Some are a genuine working budget and some are a demo. Gamma's free tier is 400 one-time AI credits, which is an evaluation. Miro's free plan includes 10 AI credits a month. Read whether the allowance renews monthly or is granted once — that single word changes what the free tier is for.
Where the numbers come from: our own hand-researched dossiers, quoted verbatim, each with the date we last verified it on the vendor's site. We publish no prices we haven't researched, and "not published" is something we say out loud. Some links are affiliate links — we may earn a commission if you sign up, at no extra cost to you, and it never changes what we write: several tools above are our partners and are criticised on this page anyway. How we review →
New dossiers, cost-traps we found, and what actually changed in AI tooling — no hype, no sponsored-disguised-as-advice. Unsubscribe anytime.