ChatGPT

Why "Just Ask ChatGPT" Is Bad Advice

May 2026

By Mark Reeves — author of all 47 AI Field Guides. Written and tested in a working business.

AI-generated portrait. Pen name. Real work.

The most common AI recommendation people give is also the laziest — and it's quietly making you worse at using these tools.

Everyone Says It

The PR manager Googled the stat after her colleague flagged it. The study ChatGPT had cited didn't exist. Neither did the journal volume. The blog post had been live for three days.

That's what "just ask ChatGPT" looks like when it goes wrong. And it goes wrong in a specific, predictable way — not because the tool is broken, but because people are routing the wrong tasks to it.

You mention you're doing research for a project. Someone says, "Just ask ChatGPT." You're struggling to write a long report. "ChatGPT can write that." You need a source for a statistic. "ChatGPT knows everything."

It's become the default answer. And it's wrong — not because ChatGPT is bad, but because people are using a drafting tool for research, a brainstorming tool for fact retrieval, and a text-completion engine for tasks that need specific, verifiable output. The tool isn't broken. The routing is.

ChatGPT has real strengths and real blind spots. Ignoring both doesn't make you efficient. It makes you frustrated and eventually convinced that "AI doesn't work."

Three Places Where This Goes Wrong

When you need real sources, ChatGPT is the wrong tool.

Ask ChatGPT to cite a study and it will often give you one. Author names, journal, year, volume number — the whole thing. The problem is that citation may not exist. The journal is real. The authors are real. The paper? Made up. This isn't a bug they forgot to fix. It's a fundamental limitation of how large language models work: they generate plausible text, not verified facts.

If you're doing research that requires real, clickable sources, use Perplexity instead. It's built to search the web in real time and surface actual links. Not perfect, but it's working from real documents — not constructing what a citation should look like.

When you need a consistent voice across a long document, ChatGPT drifts.

In my experience, consistently across different documents: give ChatGPT a 3,000-word writing task with a specific tone — formal, direct, first-person — and check section four. It starts slipping. The sentence structure gets looser. The formality drops. By the end you're editing more than the AI wrote.

Try both on your specific task and compare — but in my experience, Claude stays in character further into a long document. That's a real difference if you're doing serious long-form work.

When your prompt is vague, no tool saves you.

This one isn't about ChatGPT specifically. It's about what people actually mean when they say "just ask ChatGPT" — they mean "hand it the problem and let the AI figure out what you want." That's not how any of this works.

"Write me something about leadership" returns something about leadership. Technically correct, useless in practice. The tool doesn't know you meant a 400-word piece for a team newsletter in a tone that doesn't sound like a LinkedIn post. You have to tell it that. The AI reflects the quality of your input back at you.

The Hidden Cost

Bad AI advice doesn't just waste an afternoon. It shapes how you think about these tools.

You ask ChatGPT for a research source. It gives you a fake one. You think: AI isn't reliable. You ask it to write something long. It drifts. You think: AI can't write. You give it a vague prompt. You get a vague result. You think: AI is overhyped.

None of those conclusions are right. They're the result of using the wrong tool, or the right tool with a bad prompt. But if nobody explains that distinction, you walk away thinking the whole category is broken — and you stop experimenting.

That's the real cost. Not the wasted hour. The person who uses ChatGPT for research, gets a fabricated citation, publishes it, and gets called out — they don't try again. They write off the whole category.

What Good Advice Actually Sounds Like

Before you touch any AI tool, answer three questions: What am I actually trying to do? What kind of output do I need — a source, a first draft, a summary, a structured list? And what do I already know about this problem that I can put into the prompt?

Once you have those answers, picking the tool takes thirty seconds. Research with citations? Perplexity. Long-form writing where tone matters? Claude. General brainstorming, quick drafts, conversational back-and-forth? ChatGPT is genuinely excellent at all of that.

The tools are different. Use them differently.

What this article doesn't cover: how to tell when the right tool gave you a bad answer anyway. Both ChatGPT and Perplexity can confidently mislead you — just in different ways, at different points in the workflow. Knowing the failure mode of each tool is a different skill than knowing which tool to use.

Try This Before You Leave

Here's a two-minute test that makes the citation problem concrete. Run it yourself:

1. Ask ChatGPT for a specific statistic or study in your professional domain. Something like: "What percentage of [relevant topic] do [relevant audience] experience? Include the source."
2. Take the citation it gives you — author, journal, year — and search for it directly. Try Google Scholar or the journal's own site.
3. Now ask Perplexity the same question and compare what comes back.

You don't need to trust the article's claim that ChatGPT fabricates citations. You can verify it in the time it takes to read a paragraph. If the source checks out, great. If it doesn't, you've just confirmed exactly why tool routing matters.

This article covered which tool to use for which job. Three things it didn't cover:

What to do when the right tool gives you a bad answer anyway. Knowing tool routing doesn't protect you from citation-washing in Perplexity or from ChatGPT generating plausible-but-wrong output on questions where it should know better. There's a verification habit that closes that gap — and it's different from just "checking your sources."

The prompts that consistently produce good output across all three tools. Routing is the frame; prompting is the execution. A correctly routed bad prompt still gets you bad output.

When Perplexity's "real sources" aren't enough. Perplexity cites real documents — but summarization errors are common. The guide covers how to spot them before they reach your work.

Related reading

• How to get a useful answer out of ChatGPT every time
• The 3 mistakes beginners make with ChatGPT

Mark Reeves is a pen name. AI Field Guide publishes role-specific, practical guides for using AI tools in real work.