Are ChatGPT and friends any good at business news yet?

From shockingly bad Gemini to not bad Mistral

	Relevance	Concision	Timeliness	Sources	Overall
Perplexity	🟡 Did correctly identify important major players such as Meltwater. Failed to capture GPT Deep Research.	🔴 866 words, full of noise	🔴 Mostly ‘2026’ or 2025 and still relevant. Denies that anything has happened in the past week.	🟡 Reuters report contrasts with Meltwater’s SEO listicles.	Not really useful.
Claude	🟢 Hits multiple useful angles	🟡 Too lengthy to be ‘concise’	🔴 Didn’t grasp ‘past week’	🟢 Simon Willison’s blog + original announcement posts	Considering I’m paying, could be better.
Mistral	🟡 Goes off the rails	🟡 Some lengthy paragraphs.	🟢 (Yes this is a generous rating.)	🟡 Nothing egregious.	Given speed and underdog status, not bad
Gemini	🔴 Misses relevant, includes irrelevant AEO	🟢 At least it was wrong briefly.	🔴 ‘Code Red’ reported as past week.	🔴 Barely used sources	Shockingly bad.
ChatGPT	🟢 Verge story + inclusion of relevant arXiv papers.	🟡 Couple of bigger paragraphs	🟢 Correctly understood ‘past week’	🟡 Still duped by Red Brick’s SEO slop	Best UI and best so far.

What stood out to me

• Google Gemini was shockingly bad. But I’d noticed in the past couple of weeks Gemini 3 has become quite good. So variability is high.
• My methodology is far from perfect. Perhaps I should have put more effort into prompting.
• It stands out that only ChatGPT seems to have invested at all in UI.
• Perplexity feels like it has fallen behind.

Perplexity

Hundreds of words of regurgitated SEO noise.

I started with Perplexity.

I initially provided the following:

I work on a commercial awareness / competitive intelligence slack bot called ‘Anatole’. I want to stay on top of media monitoring, curation, recommendation systems. also keep me informed about competitor systems like Distill, and also any discussions around Google Alerts, Deep Research on a scheduled/recurring basis.

But Perplexity returned hundreds of words of noise. So I added this line.

Find me relevant articles and summarise them. Be concise.

Perplexity was not concise.

When I asked about the past week Perplexity hit me with another 342 words, again drawn from SEO blogs.

So while there’s some good stuff in there — Reuters annual report, an arXiv paper, the names of many competitors — it’s pretty unappetising.

Claude

Good until it doesn’t know what week it is.

The results are overall pretty good.

At the top we see that it pulls from Simon Willison’s ’2025: The year in LLMs’ and highlights the relevant insight that the Deep Research pattern has fallen out of fashion. That’s a ‘hit’.

Likewise, further down, the Slack agents stories.

Maybe this is just me, but I can’t trust a claim about a company when its source is a competitor’s SEO blog:

And then I asked it what happened in the last week and it became increasingly unhelpful.

First: I’m running this test on 25 January. So by the orthodox definition, things that have happened on 13 January have not happened within the last week.

Mistral

Not too shabby.

Look, I have a soft spot for Mistral. I think their Medium model is pretty good, have Le Chat in my browser sidebar because it’s so fast and, they tend to have some of the better vibes.

Even today, look at that:

I never expect it to be the best, though.

Yep, the article is from Thursday. Chapeau, Mistral.

(It does go a bit off the rails after that.)

Gemini

Bad and then you realise it doesn’t know what year it is

Having asked for articles, I get classic LLM generalities.

It starts badly. I’m fairly sure that from 19 - 25 January we didn’t shift from simple tracking to ‘Agentic intelligence’.

ChatGPT

Best UI, decent results.

I’ve criticised ChatGPT in the past but I think it’s done a decent job here.

The list of media monitoring tools is readable, correct and well-sourced.

The cards help readability and The Verge’s article is a useful one.

The Red Brick Labs SEO piece still gets through, though.

Conclusion

I’m surprised at some of these results. Perplexity has fallen out of fashion, but I would have thought that given their past news-related activity, their product would have been more tuned for news. But it was unusable.

I expected Gemini to run away with it. But on this occasion it fell into a heap.

Mistral was a pleasant surprise. Claude and ChatGPT were predictably fine.

(INSERT HARD SELL HERE)

If monitoring your competitive environment matters to you, you probably don’t want to go through this mess.

‘Oh sorry, didn’t see Acme Inc had raised. I suppose Claude was having an off day.’

anatole.fyi is tuned to ensure you never miss what matters for your work.