What You're Actually Paying For in 2026
AIサブスクリプション、中身減ってる件について
🐼You pay your AI subscriptions every month.
You're getting the snack food version of what you paid for.
Same price. Less product.
🐼Same price. Smaller product. You don't notice because the sticker didn't change.
Harder to notice because there's nothing physical to weigh.
🐼| Feature | What You Got |
|---|---|
| Price | $20/month |
| Model access | [FILL IN — e.g. GPT-4 Turbo, unlimited GPT-3.5] |
| Message limits | [FILL IN — e.g. 40 messages / 2 hours] |
| Speed | [FILL IN — e.g. priority access, no queue] |
| Extra features | [FILL IN — e.g. code interpreter, DALL-E] |
Research: check Wayback Machine for pricing page from 2024–2025
🐼| Feature | Then | Now | Change |
|---|---|---|---|
| Price | $20/mo | $20/mo | → same |
| Model access | [FILL IN] | [FILL IN] | ↓ reduced? |
| Message limits | [FILL IN] | [FILL IN] | ↓ reduced? |
| Speed | [FILL IN] | [FILL IN] | ↓ slower? |
Price stayed the same. Product shrank.
🐼$20 ÷ [X] effective uses
$20 ÷ [Y] effective uses
That's a [XX]% increase in real cost.
🐼| Feature | Then | Now | Change |
|---|---|---|---|
| Price | $20/mo | $20/mo | → same |
| Message limit | [e.g. 75 / hour] | [FILL IN] | ↓ ? |
| Model access | [Opus? Sonnet?] | [FILL IN] | ↓ ? |
| Speed | [FILL IN] | [FILL IN] | ↓ slow mode? |
Key question: has Opus been quietly replaced with Sonnet by default?
🐼| Service | Price | What Shrank? |
|---|---|---|
| Gemini Advanced | $19.99/mo | [FILL IN — what does "Advanced" mean now vs launch?] |
| Perplexity Pro | $20/mo | [FILL IN — search limits? model access?] |
| Midjourney Standard | $30/mo | [FILL IN — fast hours? quality?] |
It's not just ChatGPT and Claude. It's across the board.
🐼"Just use the free tier?" — the free tier is shrinking too.
| Service | Free Tier Then | Free Tier Now |
|---|---|---|
| ChatGPT | [FILL IN] | [FILL IN — GPT-4o-mini only? Rate limited?] |
| Claude | [5 msgs/hr?] | [FILL IN] |
| Gemini | [FILL IN] | [FILL IN — paywalled features?] |
Double squeeze: paid shrinks, free shrinks harder.
🐼Use different services for different tasks. Claude for writing, ChatGPT for coding, Gemini for research. Split your budget.
Models like Qwen 3.6, Kimi, DeepSeek, Grok — some still give you good models for free. Run locally if you can.
If you're making money with a tool, the subscription pays for itself. Casual use? Audit what you're actually getting.
You're paying the same but getting less.
The question isn't whether it's happening — it's what you do about it.
Don't put all your AI eggs in one basket.
Stop auto-renewing until you check what you're actually getting.