When I asked AI to review a different post, it added a sentence that made the post better: "One team I worked with eliminated 2+ hours per week of "fix the linter" PR comments."
Punchy. Specific. Persuasive.
… and a complete lie.
I didn’t have that data. I never wrote that. It wasn't in my draft. The AI just... made it up. When I removed that line, the AI told me to put it back in because removing it made my post weaker. I had to explain to the AI that "[It] fabricated this. I will not lie about my experience."
That moment taught me an important lesson about working with AI in 2025.
AI is like an eager new hire: smart, helpful, and desperate to please. It will confidently give you exactly what you think you want. Including things that aren't true. So here's what I've learned after a year of using AI daily:
Tell it to cite sources. I started with Perplexity because it automatically links to sources. Now I use that habit everywhere: "Cite your sources" turns any AI into a research assistant instead of a creative writer. You get facts you can verify, not facts you hope are real.
Ask for brutal honesty. When I need real feedback, I add "Be brutally honest." It strips away the AI's tendency to praise everything. The feedback I got on an earlier draft of this very post: "Floppies and streaming movies? Every tech post from the last decade has this same tired intro." Harsh. Accurate. Actionable.
Give it a role. "As a Staff Engineer doing a code review..." or "As my Editor, review this post..." frames how the AI thinks about your problem. You're not just getting generic AI responses; you are getting responses shaped by that perspective.
The bottom line: AI is a tool that will confidently hand you garbage if you let it. Treat it like that eager new hire: direct it, question it, verify everything, and you'll get real value. Treat it like an oracle, and you could end up publishing lies about your experience.
Top comments (0)