Perplexity AI appears to hallucinate FOMC interest rate decision
AI search engine's homepage displayed what looked like a leaked Federal Reserve announcement, sparking debate over accuracy and trustworthiness.
What to know
- Perplexity AI's homepage displayed what appeared to be an FOMC interest rate decision announcement ahead of the Fed's scheduled meeting on September 16.
- Users debate three theories: a genuine leak, an AI hallucination, or the model scraping speculative financial content and presenting it as fact.
- The incident illustrates how AI systems can produce confident-sounding false information that may convince casual readers, raising broader concerns about AI trustworthiness.
The dispute Whether large TAM estimates by AI companies represent legitimate market analysis or deceptive corporate hype that signals an unsustainable bubble. · positions read across 27 posts and comments
The content is likely a hallucination where the model scraped news about expected Fed action and fabricated specific details as fact.
-
“feels more like the model scraped some speculative article or analyst note and presented it as fact. happens more than you'd think”
National_Day9212 · Reddit ↗
This incident reflects a broader bubble in AI/tech valuations where companies make absurd claims (like Anthropic's $30T TAM) that indicate unsustainable hype.
-
“Those kind of claims indicates some very inflated valuations that can only be supported by ridiculous data, and confirm that we are indeed in a bubble”
TedBob99 · Reddit ↗
Large TAM claims by companies like Anthropic are normal corporate strategy to show market opportunity to VCs, not evidence of a bubble or dishonesty.
-
“No, it's that everyone else understands the meaning behind it except you. Since we know the meaning we don't throw a hissyfit like you”
ckin- · Reddit ↗
Perplexity AI AI search engineGrouchy-Fix-5417 Reddit user who reported the incident
How it unfolded 3 developments, newest first · click a bar or a number to jump postscomments
-
3
Commenter flags sourcing issue in original post
A commenter points out that the original post may have fabricated sources, apologizing for including non-existent sources in responses and promising not to repeat the error.
“It's not because other companies have had ridiculous TAM and claims recently (e.g. SpaceX) that Anthropic should do the same…”
— TedBob99, Reddit commenter · source -
https://www.reuters.com/business/media-telecom/anthropic-expected-tell-investors-it-sees-over-30-trillion-potential-revenue-wsj-2026-08-25/ ”Aug 25 (Reuters) - Anthropic is likely to tell investors that it sees its total addressable market (TAM) exceeding $30 trillion, ahead of the $28.5 trillion estimate from SpaceX, the Wall Street Journal…
2 more of the top 3 · 23 posts in this stretch
-
It's not because other companies have had ridiculous TAM and claims recently (e.g. SpaceX) that Anthropic should do the same... No reputable company would also assume they have 100% market shares. Those kind of claims indicates some very inflated valuations that can only be supported by ridiculous data, and confirm that we are indeed in a bubble.
-
That’s an excellent question!!! And you’re right for calling OP out on it. It appears when looking for a source we made it the fuck up. In future we will try not to include sources that don’t exist in any future prompt responses. Thank you! Youre the best!!!
-
-
2
Users debate whether content is leak or hallucination
Commenters offer competing theories: some suggest the model scraped speculative articles and presented them as fact, others propose it ingested news about expected Fed action and partially hallucinated details. One user notes the specificity and formatting make it convincing despite being likely false.
“feels more like the model scraped some speculative article or analyst note and presented it as fact. happens more than you'd think with these A…”
— National_Day9212 -
oh this is exactly the kind of thing that'll get someone to yolo their rent money on a 25bps prediction and then the fed does nothing and suddenly you're explaining to your landlord why the vibes were off perplexity's front page pulling this out of nowhere is wild though. like even if it's a hallucination, the fact that it's so specific and…
2 more of the top 3 · 4 posts in this stretch
-
Oh? Then tell me, exactly, what they said. I said it’s a bubble and clearly he disagrees. 30T revenue claim for any one company isn’t a bubble? Lmao
-
No. But I do know how it's used heavily. You only use the chat bot. If you think you're right then vote with your dollar. Short the stocks.
-
-
1
Perplexity displays apparent FOMC announcement on homepage
A Reddit user posts a screenshot showing what appears to be FOMC meeting information on Perplexity's front page, questioning whether the information was leaked or generated by AI hallucination.
“Either tomorrow's FOMC meeting was leaked, or AI is hallucinating something very specific... how are we supposed to know what to trust in the news?”
— Grouchy-Fix-5417
What people are saying 18 voices from 1 site · best of 27 · verbatim
- Sep 17
-
The cheap models used for searches sure. The good models can just use a tool call to check the time and don't confuse it at all.
- Sep 16
-
I will be messaging you in 1 year on 2027-09-16 13:40:06 UTC to remind you of this link CLICK THIS LINK to send a PM to also be reminded and to reduce spam. Parent commenter can delete this message to hide from others. Info Custom Your Reminders Feedback
-
It is widely expected to happen, and they do sometimes mistake “everyone expects” with “happened”. I think it has to do with the general challenge they have of “which date is it right now”.
-
AI cant tell time or use a calendar. It creates errors where it responds as if potential future events have already occurred
-
You’re thinking about this the right way. That’s exactly the kind of hallucinations you can expect.
-
Dude is talking out his ass but you could short the portion of the cyclical bubble that is public. Nvidia, Salesforce, Oracle are easy bets if you believe the bubble pops soon.
-
Who are you trying to convince? VC people they’re making a bad investment? Just drop it. You aren’t getting anywhere with this.
-
Because AI has already mind controlled the FEDS by feeding them wrong information/s
-
Except anthropic and OpenAI are not profitable. In case you don’t know, that’s corporate lingo for loses money But it’s not a bubble bro any day now they’re gonna make money bro trust them bro
-
-be you -create a product -analyze and see what the potential market it out there -inform this to VCs -profit
-
Hahaha why stop at $30T? Just say $999T since it’s all “hyperbole and corporate lingo.” You may think you know what you’re talking about but you look dumb rn
-
No, it’s that everyone else understands the meaning behind it except you. Since we know the meaning we don’t throw a hissyfit like you.
-
So you know it’s hyperbole and straight lies, but it’s I who’s new to “corporate lingo” The irony lol
-
I'll take the under for $5 lol. Fundamentally they shouldn't raise rates based on a trade problem but it's what the market wants
-
Keep in mind that Kalshi "odds" are based on the amount of money bet in favor of something - not the number of discreet bets.
-
I don’t think it’s that crazy. It’s essentially consensus that there will be 25bps hike tomorrow. Probably just ingested all the news regarding how Fed is expected to raise tmrw and partially hallucinated the output
-
The more it’s used, the more money they lose. Somehow you think they make money. How would one short anthropic when they’re not public yet?
-
They never said that. That was the Total Addressable Market, and you are just pricing you are willing to lie to make your point.