We added three tools to the directory on 3 November 2026. Reading their pricing pages back to back, each one withholds a different number — and in each case it is the number that decides whether the plan is worth buying.
None of this is fraud. All three are, in their own way, telling the truth. They are simply not telling you the one figure you came for.
1. Grok sells you a multiple without a base
Grok has five consumer tiers: Free, Plus at $8, Premium at $30, Super at $100 and Ultra at $200 a month, all excluding tax. Every tier includes what xAI calls a pool of AI usage, shared across Grok and Cursor, covering chat, image and video generation, voice and the bot.
Here is how the paid tiers describe what you get:
- Premium: “4x more usage compared to Plus”
- Super: “15x more usage compared to Plus”
- Ultra: “50x more usage compared to Plus”
Plus is never quantified. Not on the pricing page, not in the FAQ beneath it. So “50x more usage” is fifty times an unpublished number. You can work out that Ultra gives you more than Super, and that is the entire extent of what the page lets you conclude.
The free tier gets the same treatment: “Limited Chat requests” and “Limited agent requests”, with usage marked simply “Limited”.
Two credits where credit is due. The pool refreshes weekly rather than monthly, which genuinely does mean one heavy day cannot cost you the rest of the month. And xAI states clearly that API usage is billed separately at model rates and does not draw on the pool, and that third-party models inside Cursor are charged at API list price. Those are the kind of exclusions other vendors bury.
2. Lovable puts the number in the docs, not on the pricing page
Lovable’s pricing page describes the free tier’s allowance as “Free grants”. That is the whole description.
The documentation is far better. It gives the free plan five daily build credits, a 20-credit monthly Cloud grant and a 4-credit monthly AI grant, and then explains something the two-word summary on the pricing page cannot:
“After you reach the monthly cap, Lovable stops granting the daily 5 credits for the rest of that calendar month and resumes on the 1st at 00:00 UTC. For example, on Free, you receive daily build credits for the first 6 days each month (6 × 5 = 30), then none until the next month starts.”
Five credits a day, capped at thirty a month. That is six days of building, then twenty-four or twenty-five days of nothing. “5/day” is accurate and also, on its own, misleading. Lovable deserves real credit for writing the arithmetic out in full — most vendors would have left it at the cap. It is just not on the page where someone decides whether to sign up.
One more number the pricing page does not resolve: Pro at $25 and Business at $50 both include 100 monthly credits. The extra $25 buys SSO, roles and a security centre, not capacity. Lovable’s own FAQ has a question titled “Why is the Business plan more expensive?”, which suggests we are not the first to ask.
3. tl;dv shows you the annual price first
tl;dv is the mildest case, and worth including precisely because it shows what good practice looks like with one small catch.
The page opens on the annual rate: Pro at $18 per seat a month, Business at $29. Toggle to monthly and the real monthly figures appear: $29 and $39. Paying monthly costs 61 per cent more on Pro and 34 per cent more on Business.
The difference from the other two is that both numbers are published, the toggle is visible, and each price is labelled “billed annually” or “billed monthly”. Nothing is hidden. The page simply defaults to the cheaper of the two, as almost every subscription page now does. You just have to notice.
What to do about it
A pricing page is a sales document. It is written to make the next click easy, not to make your budgeting easy. Three habits cover most of it:
- Convert every relative claim into an absolute one before you pay. If a tier is sold as a multiple, find the base. If there is no base, you are being asked to buy on faith.
- Read the documentation, not just the pricing page. Lovable’s six-day arithmetic is published, honestly and in detail. It is two clicks away from the page that would have let you assume otherwise.
- Check which price the page is showing you. Annual-billed figures are now the default almost everywhere. The monthly cost is often 30 to 60 per cent higher.
And one that applies to all three: find out what happens when the allowance runs out. Grok offers top-ups and you keep going. Lovable’s documentation says that when the balance reaches zero, “building stops, AI features in deployed apps stop working, and apps that use the built-in backend (Cloud) pause” — which makes a credit balance an uptime question, not just a billing one.
What we checked
All three pricing pages and their associated documentation and terms were read on 3 November 2026 and quoted directly. Prices exclude tax unless stated. We have not tested any of the three products. Full entries, with the clauses and dates: Grok, Lovable, tl;dv.
A longer catalogue of these patterns, drawn from the whole directory, is in Eight ways AI pricing pages mislead you.