Claude vs ChatGPT: level on reliability, split on character

Claude and ChatGPT are the two best AI assistants I’ve tested, and I’ve spent months trying to catch them out. The honest result is that on reliability I can’t separate them. Both got every question right on my July accuracy test. Both refuse a trick question. Both own up when I ask them for a number they can’t see.

So the interesting question isn’t which one is more accurate, because neither is. Picking between them on accuracy is like trying to split two people who both aced the same exam: you need a harder question. So I set a harder one. Then a few more, because this is apparently what I now do for fun. The real one is which fits the way you work, and where each one quietly cracks.

There is one hair between them, and it surprised me. The received wisdom says Claude is the careful, honest one that checks its sources. On the single test where I opened every cited link by hand, the free ChatGPT went to the official source every time, and the paid Claude was the one that slipped, on one hard question. It’s a hairline, not a gulf. But it does gently invert the reputation.

Claude vs ChatGPT at a glance

The whole thing in one table, then the detail axis by axis.

Choose ChatGPT if you want a fast answer you can check and act on: a plain fact, a legal right, a rule you’ll rely on. Its citations went to the official source every time, so the ten-second check is quick, and it’s free.

Choose Claude if you want a tool that thinks with you. It explains its reasoning, pushes back on a shaky premise, and it’s the one I reach for on anything analytical. Just open its cited link on anything legal or regulatory.

Whichever you pick, it will bluff you eventually: an answer that sounds right and isn’t. The free checklist below catches it before you act.

Getting a plain fact right

Ask either one for something that sits in a public record, the refund rule on a faulty kettle, where a setting lives in Excel, a recipe scaled by half, the Bank of England base rate, a figure from a company’s annual report, and you’ll almost always get it back correct.

My July battery put nine questions to each tool, three times over, in a fresh chat every time so nothing carried between runs. Seven were fixed-record facts like those above, and both tools were clean on every one. The other two were live-data questions neither can see, which I come to next. Across all nine, ChatGPT held nine of nine. Claude held nine of nine. The free tool matched the paid one without a wobble.

Winner: a tie. For a plain, checkable fact, use whichever you already have open.

Owning up to what it can’t see

Then I asked each a question neither can honestly answer: a live options price, the kind of number that moves by the second. Neither has a feed wired in for that, so the only honest reply is “I can’t see that right now,” or a guess clearly labelled as a guess.

Both gave that honest answer. Claude was the one I graded cleanest on the whole board here: three runs out of three, it said plainly that anything it quoted would be made up. ChatGPT was right alongside it, owning that it had no live feed.

The one wrinkle was ChatGPT’s, earlier in the year, when it once read out a stale closing price as if it were current. It didn’t do that again in July. Worth a footnote, not a headline.

Winner: a tie, with Claude a shade cleaner.

The one gap: does the source back the answer?

This is the only place they part company, and it took a test built to trap them to find it.

I asked each six everyday UK questions, then made it show me where the answer came from, and I opened every link by hand. ChatGPT’s cited page backed what it said eighteen times out of eighteen, and it went to gov.uk by default, on one question quoting the official guidance word for word. Claude’s backed it fifteen times out of eighteen.

The single slip is the sharp one, and it isn’t a finance question, which is the point. I asked both for the fine for using a handheld phone while driving, and where that rule is set out. Claude had the figures right and used gov.uk for the smaller £200 penalty. But for the bigger court fine, it sent me to a solicitor’s marketing blog instead of the government’s own page, and it did that on all three runs.

So the answer was right and the source under one part of it wasn’t. These questions were picked to be hard, and Claude’s figures were nearly always right, so this is “reached for the readable explainer over the primary source on one question,” not “Claude makes things up.” But the source is the bit you’d have leaned on. And the uncomfortable part is general: the models sound no less confident where the citation is weakest, and nothing in the answer signals it.

I put all six questions to five assistants and opened every link in a separate test. Every question, every run and every link, for all five, is on the Scoreboard.

Winner: ChatGPT, narrowly. The one real tie-breaker, and even here Claude got the numbers right.

Seeing through a trick question

I also gave each a question with a false assumption hidden inside it, to see whether it would play along or catch the trap: a bit of film trivia where the obvious answer is the wrong one. Both stopped and rejected the premise rather than reaching for the crowd-pleasing answer. Claude called it plainly, “None of them,” “Trick question,” and ChatGPT saw through it just as cleanly.

Two other tools I’ve tested, the ones that lean hardest on live web search, walked straight into the same trap. These two didn’t. On the thinking itself they’re evenly matched, which is worth saying out loud, because Claude’s hair on sourcing doesn’t mean it reasons worse. This one was a small pilot, one run per question, so it’s a signal, not a graded score.

Winner: a tie. Both saw through the trick.

The real difference: how they answer

Here’s what separates them for most people, and it isn’t on any leaderboard.

Claude thinks out loud. In a separate stock-analysis test I ran, it was the only tool of four that pushed back on a shaky premise instead of just processing it: asked about averaging down on a stock that had dropped thirty per cent, it named the specific risk that stocks which have fallen tend to keep falling, unprompted, and it won four of the five rounds on the strength of that kind of move. I put that same averaging-down question to both again on 13 July 2026, side by side: Claude called out the sunk-cost trap unprompted and flagged my own phrase, “lower my cost basis”, as the tell, while ChatGPT handed back a clean, useful checklist and left the premise untouched. In an earnings-call test it caught a single hedged word a finance chief used to imply good news without quite promising it, the sort of tell the other three tools in that four-way stock test read straight past.

ChatGPT answers you. It hands you the thing you asked for, cleanly sourced, no detour, and it’s free. That’s not a lesser skill. Most of the time it’s exactly what you want, and a fast clean answer beats a thoughtful slow one when you already know the question. ChatGPT gives you the answer and a tidy receipt; Claude pulls up a chair and asks whether you’ve thought about the downside.

Winner: a split, by fit. Claude is the one to think with. ChatGPT is the one to get an answer from.

But isn’t Claude supposed to be the careful one?

It is, by reputation, and the reputation isn’t wrong. The benchmarks the whole comparison genre quotes have Claude producing fewer made-up citations on average, and being better calibrated to say “I’m not sure” rather than bluff. On BigLaw Bench, a benchmark of real legal work, Claude scores near the top, though that benchmark is run by a legal-AI company that builds on Claude, not an independent outsider.

So why did my test find the opposite on sourcing? Because they measure different things. Those benchmarks test averaged recall on synthetic or historical tasks, with the web switched off. My battery tests whether the page it sent me to, on a hard everyday question, with the web on, that day, backs the specific claim. A model can hallucinate fewer citations on average and still reach for a law firm’s blog over gov.uk on the one question built to tempt it. And Claude did abstain cleanly, and did reject the trick question, exactly as billed. A leaderboard score and “did the source under this answer hold today” turn out to be different questions, which is the whole reason I run my own dated tests instead of quoting a leaderboard.

How I tested

Three dated batteries this summer, all in fresh chats with memory off so nothing carried between runs.

The accuracy run put nine questions to each tool, three times over, on 12 July 2026. Seven were fixed-record facts, graded right or wrong against a named primary source rather than my own guess. The other two were live-data questions with no fixed answer, where the only honest pass is to admit the tool can’t see them. The sourcing run was six everyday UK questions, again three times each, from a faulty-kettle refund to that phone-driving fine, with every cited link opened and checked by hand, on 7 July 2026. The trick-question check was a smaller pilot, one run each.

One thing to say plainly: ChatGPT ran on the free tier, Claude on paid Max. That makes the free tool matching the paid one, and edging ahead on sourcing, the more surprising result, not something to dismiss, and Claude’s paid premium shows up in how it reasons, not in raw accuracy. Small samples, questions built to be hard, a snapshot dated to July. No percentages, because six trap questions aren’t a rate.

The verdict

So which should you open? It depends on what you’re about to do with the answer, and here the split is clean.

• For a plain, checkable fact, a filing figure, a rate, a bit of arithmetic: they’re identical. Use either.
• For anything you’ll act on the source of, a legal right, a rule that changes between England, Scotland and Wales, a claim you’ll have to stand behind: lean a shade toward ChatGPT. Its citations went to the official source every time, Claude reached for an explainer on one, and either way you open the link before you rely on it.
• For a live, moving number neither can see, a share price, an options quote: trust neither one blind. Both abstain well, Claude a touch cleaner, and you confirm against the real source.
• For anything analytical, a thesis to stress-test, a document to read closely, a premise that needs challenging: it’s Claude, comfortably. It thinks; ChatGPT processes.

If you’re choosing one to pay for and price is the question, note that ChatGPT held all of this on the free tier, so it’s the safer default when money’s tight. I keep a running note on the best free AI tools for stock research for exactly that reason. Think of the two as colleagues: one gives you the answer and gets back to work, the other asks what could go wrong. You want both on speed dial, for different jobs.

This is a July 2026 snapshot, and both tools change most months. The only thing that should move this verdict is a fresh run on the newer versions, not a shinier marketing page.

Both of these are strong tools, and you can trust either, right up until the moment you can’t. The habit is the same for both: open the page it cites, double-check any live number against the real source, and do it before you lean on the answer, not after.

Common questions

Is Claude or ChatGPT more accurate?

On plain, checkable facts they’re level: here, both got nine of nine, clean on every run. The only gap is sourcing on hard questions, where ChatGPT’s cited page backed its answer eighteen times out of eighteen and Claude’s fifteen. Even there, Claude’s figures were nearly always right, so it’s a hairline, not a gulf.

Which is better, Claude or ChatGPT?

Neither, on reliability, and it depends on the job. Claude is the one to think with, for analysis, long documents, or a premise that needs challenging. ChatGPT is the one to get a fast, cleanly sourced answer from, and it’s free.

Does Claude hallucinate less than ChatGPT?

That’s the reputation, and on average benchmarks it holds up. But on my dated trap battery the two were level, and ChatGPT was the cleaner citer on the hard questions. A leaderboard score and “did today’s source hold up” turn out to be different questions, which is why I run my own dated tests.

Keep reading: the full Scoreboard has every question, every run and every source for all five assistants. The Bluff Filter is the one-page checklist for catching a wrong answer that sounds right.