Can’t you just paste it all into ChatGPT?
You can — for one draft. The trouble is everything around the draft: a chatbot guesses who exists, can’t confirm who’s currently where, has no real ranked index, and can’t send, track, or follow up. Here’s the honest difference.
Generates plausible-sounding names from training data. It can't enumerate a live, ranked index and can't confirm who is currently at which institution — so it invents people, or lists ones who have moved or retired.
Pulls real professors from a live academic index (OpenAlex), resolves and confirms current affiliation, and ranks by recent publishing output.
You paste your resume and interests again every session. The match is vibes — it has no grounded notion of research overlap.
Resume-grounded matching by genuine research overlap, with a sourced 'why this match' rationale for each professor.
Citations are frequently fabricated or out of date — a known failure mode for any LLM asked to recall specific papers.
Every claim links to a real paper URL and the professor's faculty page. Nothing is shown without a source.
Tends to produce the openers, declared passion, and flattery that PIs skim past and recognise as AI-written.
Drafts in your voice and flags AI tells before you send, so the email reads like a person wrote it.
Can't send, can't track. You copy text out and manage everything by hand.
Approval-gated Gmail send — you approve every email, and Marrow never reads your inbox.
Stateless. It forgets every professor between chats, so you rebuild context each time.
A tracked pipeline — contacted, awaiting reply, replied, follow-up due — so you never lose the thread.