AILIMIT.WATCH
LIVE

AI Shrinkflation Index

How often AI vendors quietly tighten what you pay for — usage limits, quotas and pricing. Graded on how often they tighten over time, not on product quality. Worst-first.

Powered by the AILimit Stability Index · 50% limit · 30% pricing · 20% transparency · trailing-12-month weighted · 30 vendors

01 / THE INDEX
VendorOvrLimPriTrn
Perplexity
Rated
D
F
A
F
Limit 3 tightenings · Feb 26 (silent) · Feb 26 (silent) · Feb 26 (silent)
Pricing — no pricing tightening recorded in 11mo
Transparency 3 tightenings · Feb 26 (silent) · Feb 26 (silent) · Feb 26 (silent)

Model access (informational): A · excluded from Overall

Gemini
Rated
B
C
A
A
Limit 2 tightenings · May 26 · Nov 25 (silent)
Pricing — no pricing tightening recorded in 8mo
Transparency 1 tightening · Nov 25 (silent)

Model access (informational): A · excluded from Overall

GitHub Copilot
Rated
B
B
B
A
Limit 2 tightenings · Jun 26 · Jun 25
Pricing 1 tightening · Jun 26
Transparency — no transparency tightening recorded in 12mo

Model access (informational): D · excluded from Overall

Claude
Rated
A
B
A
B
Limit 2 tightenings · Mar 26 (silent) · Aug 25
Pricing — no pricing tightening recorded in 14mo
Transparency 1 tightening · Mar 26 (silent)

Model access (informational): B · excluded from Overall

Replit
Rated
A
A
B
A
Limit 2 tightenings · Apr 23 · Oct 22
Pricing 4 tightenings · Feb 26 · Sep 25 (silent) · Jul 25
Transparency 1 tightening · Sep 25 (silent)

Model access (informational): A · excluded from Overall

Cursor
Rated
A
A
A
A
Limit 2 tightenings · Nov 25 · Sep 25
Pricing 2 tightenings · Jun 26 · Jun 25
Transparency — no transparency tightening recorded in 12mo

Model access (informational): A · excluded from Overall

Notion AI
Rated
A
A
B
A
Limit 1 tightening · Mar 26
Pricing 2 tightenings · May 26 · May 25
Transparency — no transparency tightening recorded in 40mo

Model access (informational): A · excluded from Overall

v0
Rated
A
A
A
A
Limit 1 tightening · Apr 25 (silent)
Pricing 2 tightenings · Jan 26 · May 25
Transparency 1 tightening · Apr 25 (silent)

Model access (informational): A · excluded from Overall

Luma Dream Machine
Rated
A
A
A
A
Limit 1 tightening · Nov 24 (silent)
Pricing 2 tightenings · Sep 25 · Nov 24
Transparency 2 tightenings · Nov 24 (silent) · Nov 24 (silent)

Model access (informational): A · excluded from Overall

Sora
Rated
A
A
A
A
Limit 2 tightenings · Nov 25 · Oct 25
Pricing — no pricing tightening recorded in 18mo
Transparency — no transparency tightening recorded in 18mo

Model access (informational): A · excluded from Overall

Windsurf
Rated
A
A
A
A
Limit — no limit tightening recorded in 13mo
Pricing 1 tightening · Mar 26
Transparency — no transparency tightening recorded in 13mo

Model access (informational): A · excluded from Overall

Devin
Rated
A
A
A
A
Limit — no limit tightening recorded in 18mo
Pricing 1 tightening · Apr 26
Transparency — no transparency tightening recorded in 18mo

Model access (informational): A · excluded from Overall

Suno
Rated
A
A
A
A
Limit — no limit tightening recorded in 24mo
Pricing 1 tightening · Nov 25
Transparency — no transparency tightening recorded in 24mo

Model access (informational): A · excluded from Overall

Midjourney
Rated
A
A
A
A
Limit — no limit tightening recorded in 12mo
Pricing 1 tightening · Jun 25
Transparency — no transparency tightening recorded in 12mo

Model access (informational): A · excluded from Overall

Synthesia
Rated
A
A
A
A
Limit 1 tightening · Jun 24
Pricing — no pricing tightening recorded in 29mo
Transparency — no transparency tightening recorded in 29mo

Model access (informational): A · excluded from Overall

Tabnine
Rated
A
A
A
A
Limit — no limit tightening recorded in 25mo
Pricing 1 tightening · Nov 25
Transparency — no transparency tightening recorded in 25mo

Model access (informational): B · excluded from Overall

ChatGPT
Rated
A
A
A
A
Limit 1 tightening · Aug 25
Pricing — no pricing tightening recorded in 10mo
Transparency — no transparency tightening recorded in 10mo

Model access (informational): A · excluded from Overall

Grok
Rated
A
A
A
A
Limit — no limit tightening recorded in 11mo
Pricing — no pricing tightening recorded in 11mo
Transparency — no transparency tightening recorded in 11mo

Model access (informational): A · excluded from Overall

DeepSeek
Rated
A
A
A
A
Limit 1 tightening · Nov 24
Pricing — no pricing tightening recorded in 19mo
Transparency — no transparency tightening recorded in 19mo

Model access (informational): A · excluded from Overall

ElevenLabs
Rated
A
A
A
A
Limit — no limit tightening recorded in 12mo
Pricing — no pricing tightening recorded in 12mo
Transparency — no transparency tightening recorded in 12mo

Model access (informational): D · excluded from Overall

Mistral Le Chat
Rated · quiet
A
A
A
A
Limit — no limit tightening recorded in 28mo
Pricing — no pricing tightening recorded in 28mo
Transparency — no transparency tightening recorded in 28mo

Model access (informational): A · excluded from Overall

Character.AI
Rated
A
A
A
A
Limit — no limit tightening recorded in 37mo
Pricing — no pricing tightening recorded in 37mo
Transparency — no transparency tightening recorded in 37mo

Model access (informational): A · excluded from Overall

Kling AI
Rated · ⚠ low-signal
A
A
A
A
Limit — no limit tightening recorded in 23mo
Pricing — no pricing tightening recorded in 23mo
Transparency — no transparency tightening recorded in 23mo

Model access (informational): A · excluded from Overall

Ideogram
Rated
A
A
A
A
Limit — no limit tightening recorded in 25mo
Pricing — no pricing tightening recorded in 25mo
Transparency — no transparency tightening recorded in 25mo

Model access (informational): A · excluded from Overall

Insufficient history — not graded
HeyGen
Provisional
↘ tightening
Poe
Provisional
→ stable
Lovable
Provisional
→ stable
Bolt.new
Provisional
→ stable
Leonardo.ai
Provisional
→ stable
Runway
Newly tracked

How to read this

  • Grades (A–F) are shown only for Rated vendors — those with enough sourced history. A high grade means few recent restrictions in our record, not product quality.
  • Quiet = Rated but no changes recorded in 12 months (stable by absence, not by recent observation).
  • Provisional / Newly tracked get a direction signal, not a letter — too little history to grade fairly.
  • ⚠ low-signal = we can't reliably detect new changes for this vendor (e.g. a client-rendered pricing page); its grade reflects history only.
  • Every grade derives from dated, sourced events — expand a row to see them. Not a scientific sample.
  • Full grade definition (bands, weights, coverage tiers): methodology →