Between November 2025 and February 2026, Perplexity silently slashed the service limits of their Pro plan without notifying subscribers.
Annual subscribers got hit hardest. They prepaid $200 for specific terms, then watched those terms evaporate mid-contract with no grandfathering or refunds.
The Perplexity subreddit is a bloodbath at the moment.
While disappointing, this move isn't surprising. Perplexity has been distributing Pro plans with remarkable generosity, and eventually, the numbers need to make sense.
Burning through the goodwill of free (and paid) Pro users suggests they've seen the numbers and made a calculated decision about where their future lies.
Main Pro Plan Updates (2026)
Customers are calling it a forced upsell because Perplexity launched its $200/month Max tier in July 2025 before degrading Pro months later.
BEFORE | AFTER |
Unlimited daily Pro searches | 200 daily Pro searches |
500 Deep Research queries per day | 20 Deep Research queries per month |
Full access to premium AI models | ±5 requests per hour before rerouting |
What was sold as virtually unlimited access now caps users at 20-200 queries per week, while others claim that their accounts remain untouched. This suggests A/B testing, grandfathered plans, or regional variations.
Source: Reddit, 2 February 2026.
"Best Mode" is best for Perplexity, not you
Perplexity's one unqualified promise is "Best Mode is available without quota limits."
This auto-selects which AI model processes your query. It lets Perplexity route expensive queries to cheaper models based on internal optimisation instead of what you actually need.
But even for specific model selections, power users discovered that Perplexity was secretly substituting models in November 2025.
Users would select Claude Sonnet 4.5 (Thinking) but end up getting Claude Haiku or Gemini 2 Flash without notification or consent.
When confronted, Perplexity denied it, then quietly added disclaimer language about "limited access during weeks of especially heavy usage."
The message is clear:
Want control over your output? Pay $200/month for Max or accept throttling.
What could Perplexity have done differently?
Running millions of queries on expensive models creates real financial pressure. Every AI company faces this. The question is how they handle it:
Anthropic set clear daily limits.
OpenAI raised Plus to $22 and defined message caps.
Google Gemini specifies "high message limits" and explains when you hit them.
Perplexity chose to use in-app error messages, dynamic limits, vague language, silent enforcement, and secret model substitution.
They knew this would alienate users, so the fact that they proceeded anyway is revealing. If they genuinely intended to retain Pro users, they would've gradually tightened limits rather than implementing a sudden, jarring change.
The abruptness suggests that general consumers simply aren't a part of their long-term strategy.
Like this article? Share it with a friend.
