For PhD researchers & early-career academics
Most researchers are stuck between two bad options: avoid AI and fall behind, or use it quietly and hope nobody asks. There is a third option, and it is learnable. I coach you to build an AI-assisted research workflow you could explain, in detail, to your supervisor, your examiners, or a journal editor.
No obligation. If I'm not the right fit, I'll tell you on the call.
Written by the author of AI in Academia: A Practical Guide.
You've read the misconduct cases. So you don't touch it, while colleagues quietly move through literature three times faster, and you lose months to work a tool could have scaffolded in an afternoon.
You use it, but you're not sure where the line is. You couldn't fully explain your process if asked, and there's a low hum of worry every time you submit something.
Both are expensive. The first costs you time you don't have. The second costs you sleep, and occasionally careers. The problem isn't the tools. It's that nobody taught you the method.
Where legitimate assistance ends and misconduct begins, grounded in your institution's policy and your target journals' author guidelines, not in generic advice from the internet.
Concrete, tested methods for literature discovery and screening, note synthesis, coding and analysis support, drafting, and revision, with the traps named for each.
Lightweight documentation and disclosure habits, so if anyone asks what you did, you have an honest, complete answer ready, and you knew it was ready in advance.
Models change every few months. The evaluation habits do not: verifying claims, catching fabricated citations, knowing what to never delegate.
Tell me where you are: discipline, stage of candidature, what you've already tried, what worries you. I'll tell you honestly whether coaching helps, and what I'd focus on.
Your university's AI policy, your department's norms, your funder's and target journals' rules. These genuinely differ, and most advice online ignores that entirely. We work from your documents, not from assumptions.
Not toy examples. Your literature, your data, your chapter. You leave each session with something working, plus the reasoning behind it so you can adapt it later.
60 minutes
One focused problem. Best when you have a specific question: a stage of the process, a tool decision, or a policy you can't interpret.
Most researchers start here
4 sessions
End-to-end. We work through discovery, synthesis, analysis support, and writing, building and testing your workflow on live chapters as we go.
Workshop
For supervisors advising students, or a lab agreeing shared norms. Run as a workshop for your group, tailored to your department's policy.
All sessions are online and one-to-one unless arranged otherwise. Start with the free fit call. Pricing is discussed there, with no pressure.
Alex Xavier, PhD
Author, researcher, and coach
I work at the intersection of research practice and AI tooling, helping researchers adopt these tools with the same rigour they'd apply to any other part of their method.
My position is simple: AI assistance is not inherently misconduct, and pretending otherwise has left a generation of researchers to figure it out alone, badly. What matters is knowing what you delegated, verifying what came back, and being able to say so plainly. That is a skill, and it can be taught.
No. And if that's what you're after, I'll decline the work. Coaching covers legitimate assistance: finding and screening literature, interrogating your own arguments, improving clarity of writing that is yours, and supporting analysis you understand and can defend. It explicitly does not cover generating research contributions you then present as your own, or hiding AI involvement from anyone entitled to know about it. The point of the work is to make your process more transparent, not less.
We read it together and work out what it actually commits you to, including where it is genuinely silent, which is common. Where it's ambiguous, I'll help you draft the question to put to your supervisor or research office in writing, so you get a documented answer rather than a corridor conversation. I can't give you permission your institution hasn't given, and I won't pretend to. Getting that in writing protects you.
These are moving, and they differ by publisher. Most now require disclosure of AI use in preparation, and essentially all refuse to list an AI system as an author. We check the specific author guidelines for the venues you're targeting and build your disclosure habit around the strictest one you're likely to meet, so you're never caught out by a submission form asking a question you hadn't considered.
The method transfers; the specifics change a great deal. What helps a historian working with archival sources differs sharply from what helps someone running experiments or doing qualitative coding. Tell me your field on the fit call and I'll be straightforward about where I can help most and where you'd be better served elsewhere.
Whichever ones serve the task, including free ones. Where a tool doesn't earn its place, I'll say so. I'm not affiliated with any vendor and take no referral commissions. More of the value is in method and judgment than in any particular product, which is deliberate: the tools will have changed by the time you submit.
In my experience supervisors object to concealment far more than to the tools. Several researchers have used their written AI-use protocol to open exactly that conversation and found it went well. If it would help, the protocol is written to be shown.
Yes. Unpublished work, data, and anything about your institutional situation stays between us. I don't use your material as examples anywhere without asking you first, in writing.
Twenty minutes, free, no obligation. Bring the thing you're most unsure about.
Book your fit callPrefer email? Write to hello@metascholar.co and tell me where you are in your research.