If you're applying to an AI company and not using AI tools in your application process, you're sending the wrong signal. These companies build AI products. They want to hire people who know how to use them well. But "use AI" doesn't mean "have ChatGPT write your cover letter." The bar is higher and more specific than that.
Anthropic's position: use Claude, but show your thinking
Anthropic has been the most explicit about this. Their job postings and recruiter guidance actively encourage candidates to use Claude during the application and interview process. The reasoning is straightforward: if you're applying to work at the company that builds Claude, you should be fluent in using it.
But Anthropic's expectation isn't "let Claude do the work." It's "use Claude the way a skilled practitioner would." That means:
Using Claude to research the role and company. Pull up their recent blog posts, technical papers, and product announcements. Ask Claude to help you identify themes, synthesize what the team is focused on, and spot connections to your own experience. This is the kind of prep that used to take hours. Now it takes minutes, and not doing it reads as not caring.
Using Claude to pressure-test your own writing. Draft your cover letter yourself, then ask Claude to identify weak spots: vague claims, unsupported assertions, generic phrasing. Use it as an editor, not a ghostwriter. The difference is obvious to anyone who reads applications all day.
Using Claude in take-home exercises. Many AI companies give take-home projects. Anthropic expects you to use Claude here, and they evaluate how you use it. Did you break the problem down well? Did you prompt effectively? Did you catch errors in Claude's output? Did you iterate? The meta-skill of working with AI is what they're actually testing.
What they don't want: a cover letter that reads like a prompt output. Flat, correct, generic, hitting every keyword but saying nothing real. If your application sounds like it could belong to any candidate for any AI PM role, Claude wrote it and you didn't edit it enough.
OpenAI and Google: similar signals, less explicit
OpenAI hasn't published formal guidance the way Anthropic has, but their hiring managers have said publicly (on podcasts and in blog posts) that they expect candidates to be comfortable with AI tools. Several OpenAI interview loops include live sessions where candidates are encouraged to use ChatGPT. The evaluation isn't about whether you got the "right answer" but whether you demonstrated fluency in prompting, iteration, and critical evaluation of outputs.
Google's AI teams (DeepMind, Google AI) take a more traditional approach to interviews, but their job descriptions increasingly list "experience using LLMs in production" or "familiarity with generative AI tools" as requirements. The expectation is that you've used these tools in real work, not just for demos.
The right way to use AI in your job search
Research, don't regurgitate
Use AI to understand the company, the team, and the role better than you could from the JD alone. Ask it to analyze the company's recent product launches, identify their technical challenges, and find connections to your background. Then write about those connections in your own words.
Edit, don't generate
Write your resume bullets and cover letter yourself first. Then use AI to tighten the language, catch redundancy, and flag where you're being vague. You should sound more like yourself after editing, not less.
Show the collaboration
In take-home exercises or portfolio pieces, include a brief note about how you used AI tools. "I used Claude to explore edge cases in the eval framework" or "I used GPT to generate test data, then validated the output against real production logs."
Know the limits
AI tools hallucinate, miss context, and produce generic outputs when you give them generic inputs. If you can mention a time you caught an error in an AI output and adjusted your approach, that lands better than a polished but uncritical workflow.