How honest they are.

The price of extra credits is often hidden

Of the 133 plan tiers that sell a top-up, only 61% publish what it costs. For the other 39% the top-up price surfaces only at the in-app purchase screen — after you’ve subscribed, where no crawler and no buyer comparison-shopping can see it.

Incumbents peg the credit to a cent

12 of the 51 products tie one credit to a fixed dollar value rather than a variable amount of compute. The sharpest convergence is among incumbents bolting a meter onto an existing seat product: GitHub Copilot, Atlassian Rovo, monday.com, and HubSpot have each independently landed on 1 credit = exactly $0.01. A round cent is easy to reason about and easy to reprice — the catch: when the peg holds but the credits an action costs goes up, the price rises with nothing on the page changing. That credit burn-rate repricing is tracked under What’s changing.

The opacity watchlist

And that’s only the products that publish a pricing page at all. These category-leading vertical and enterprise AI products publish none. They fail the inclusion test — and that failure is the finding. Where a pricing model is listed it is press-derived, not observed, and labelled so.

• Abridge — Ambient clinical documentation (AI medical scribe) · Per-clinician enterprise licensing, quote-only via health-system procurement (press-derived: Contrary Research; KLAS 2024)
• Ada — Customer service / CX agents · Enterprise quote-only, usage-based tied to automated resolutions (press-derived; circulating per-resolution figures are third-party and unverified)
• EliseAI — AI leasing/operations automation for housing (+ healthcare admin) · None publicly known — quote-only; company blog discusses cost drivers without numbers
• Forethought — Customer service / CX agents · Hybrid platform fee + outcome-based usage, quote-only (stated on Forethought's own pricing page, no numbers)
• Glean — Enterprise work AI / search and assistant · SPECIAL: 'Enterprise Flex' hybrid — per-user seat fee + pooled FlexCredits with a PUBLIC numeric credit burn-rate card (official docs: Fast Mode query ~3 credits p50, Thinking ~7, premium ~35, slides ~45, deep research ~33) but NO public dollar prices
• Harvey — Legal AI (large law firms + enterprise legal) · Custom enterprise contracts; per-seat figures circulate on competitor/aggregator blogs only — not citable
• Legora — Legal AI (M&A diligence, contract review) · None publicly known from primary sources
• OpenEvidence — Clinical evidence search / medical Q&A · SPECIAL: free for verified US clinicians (NPI-gated), ad-supported by pharma/med-device advertisers — no paid plans (press-derived: NBC News; corroborated by company PR)
• PlayAI (formerly Play.ht) — Text-to-speech / AI voice · Historically character-metered (Free ~12,500 chars/mo; 'Unlimited' ~$49/mo with a 2.5M-chars/mo fair-use cap) per the old play.ht site — model status post-rebrand unconfirmed
• Relevance AI — AI workforce / agent platform · Inverted opacity: the METER is public (official docs publish allocations and top-up rates — Actions $80/1,000; Vendor Credits $20/10,000, rolling over indefinitely while subscribed) but NO plan subscription price appears anywhere public
• Rogo — AI for investment banking / institutional finance · Enterprise subscription via multi-year direct contracts; no numbers from company or primary press
• Sierra — Customer service / CX agents · Outcome-based, pay-per-resolution (company's own homepage + blog 'Outcome-based pricing for AI Agents')
• Writer — Enterprise generative AI / content platform · Per-seat subscription with 'fixed credit limits' acknowledged in the plans-page FAQ — but neither prices nor meter sizes are public (historically the page showed per-seat prices with word limits; both pulled)

How opacity stacks across the panel — and which products combine hidden top-ups, contact-us walls, and variable burn — is scored on The big picture.