I'm Alice Hunsberger. Trust & Safety Insider is my weekly rundown on the topics, industry trends and workplace strategies that Trust & Safety professionals need to know about to do their job.

This week, I'm sharing how and why I use AI for writing, and when I don't use AI. To be honest, we probably should have published this a while ago (I think disclosing this information is important), but better late than never.

I know it's easy to gloss over the big red box half way down every newsletter, but I have a special request for job searchers this week.

If you want to update your newsletter settings for T&S Insider or Week in Review, you can do so via Your Account. And you know where to get me if you have any questions about today's edition. Here we go! — Alice

The Haidt of Hypocrisy - Ctrl-Alt-Speech

In this week’s round-up of the latest news in online speech, content moderation and internet regulation, Mike and Ben cover:Jordan escalates global tech argument, with Farage’s help (Politico) Farage’s rules for free speech: talk about anythi…

Buzzsprout

How I use AI for writing

In the year and half since I’ve been writing T&S Insider, I’ve experimented with using AI for writing in a variety of ways. While I wouldn't say I'm reliant on it, it's certainly part of my regular workflow now. So, inspired by other newsletter writers that I admire, I wanted to share how AI helps me write.

The disclosures

Every opinion I share here at T&S Insider is 100% my own, meaning I don't write about or share any links because someone asked me to. But, like everyone, I have some priors that are important to note in a conversation about AI.

Between October and March 2023, I was a freelance adversarial/red team tester for OpenAI prior to GPT-4’s release. 

Since March 2025, I have been Head of Trust & Safety at Musubi, where we use frontier AI models for Trust & Safety work. Part of my role includes marketing/ content creation about the tools we build and how AI can be used for Trust & Safety. 

Going forward, I will remind T&S Insider readers about this at the top of newsletters in which I talk about AI.

My approach

In general, I find AI to be a helpful tool, but I always keep in mind why I write at all:

• To share the perspectives of Trust & Safety practitioners, especially in areas that I feel are underrepresented or not discussed at all. These ideas have to come from me — the whole point is that they’re my personal perspective.
• To improve and sharpen my own writing skills and ability to make a good argument. I won’t get better at writing if I over-rely on AI.
• To share content in a way that isn’t controlled by the whims of social media algorithms, where I can build on my personal “brand” and potentially open up exciting opportunities for the future. This is only possible if people respond well to what I’m writing and it feels authentically mine.

That said, I spend a lot of time writing (both for work and for T&S Insider) and I am not a trained or professional writer. So, while I want to be sure to keep my authentic voice and perspective when writing, I can use all the help I can get. 

What I use

Tools I pay for and regularly use

• Claude Sonnet 4 - I personally pay for a pro subscription for access to Anthropic's Claude. It has high EQ and I find the UX to be the most pleasing. I find the model good for critiquing article drafts as a persona, and I use it frequently for miscellaneous personal stuff.
• Gemini 2.5 Pro and custom Gems - Included with my work's Google plan, I find Gemini good for work first drafts, LinkedIn outlines, grammar/brevity checks and summarising data. It's a total workhorse when combined with the G-Suite of tools for work (mail, calendar, etc). Nano Banana is very cool for making silly photos.
• Lex - Helpful for brevity checks and removing filler words (which I’m always guilty of using). I don't currently pay for it.

If you're interested in which AI models are the "least worst" in terms of exploitation, environmental issues, etc, this guide is helpful.

What I don’t use AI for

Things I've tried and stopped, or never did at all.

• Story ideas - Ideas are always generated by me and/or Ben. I’ve tried asking AI to generate ideas, but results are generic and overdone (which makes sense– AI is trained on already-written content, after all). I once pasted my past article titles into Claude to see if it could suggest more ideas that weren’t repetitions of what I’d already covered, and it wasn’t able to, replying: “This might actually be a sign that you've exhausted the core T&S topics from your insider practitioner perspective.” I disagree! There’s always more to write about, and that’s what makes T&S so exciting. 
• Full drafts - Asking AI to create a full draft based on my own outline rarely outputs an article that is anywhere close to meeting my standards: there are mistakes, repeated ideas, or sections that are far too short or far too long. I end up going back and forth with the AI so much that it doesn’t save me time. 
• Fact checking - I don’t rely on AI to fact-check my own work. I also don’t trust any citations or claims in AI summaries or outlines without cross-checking other sources myself. 
• Professional editing - Ben always does a professional edit before sending the newsletter out. Some articles are much more work than others, but Ben always manages to make the articles feel like me, even when he heavily rearranges things. I like having AI critique my drafts, but I love having a human editor.
• Writing for social media - I’ve played around with using AI to draft LinkedIn posts based on my T&S Insider articles. Even though I’ve inserted my own posts to a custom Gem so it can copy my style and tone, the output is never quite right. If I’m feeling stuck on what to highlight, I’ll ask AI to pull out the main 2-3 points from each article, and use that as a jumping off point for LinkedIn posts. Usually the most authentic and silly LinkedIn posts are the ones that do the best, anyway.

What I do use AI for

• Structuring and organising - I often ask AI to organise my rough thoughts into a structured format or outline. This can include suggestions for how to begin or end an article as well as overall order of ideas.
• Critiquing first drafts - I have a habit of defaulting to writing for T&S practitioners like myself, but EiM readers come from a wide variety of backgrounds, so I’ll ask AI to critique my draft based on other reader personas. I’ll ask things like “what am I missing?” or “what else should I consider based on ___ perspective?” I often disagree with what the AI suggests, but the simple exercise of going back and forth is useful. Sometimes during this process, I will use AI to generate re-writes of certain sections based on what the critiques have uncovered. I always edit the writing, grammar, and phrasing so that it continues to match my style and convey my ideas.
• First pass copy editing - Before sending to Ben, I ask AI to find gaps in my arguments, repeated phrases, unclear sections, unnecessary words, confusing sections, unexplained jargon, etc. I'll also AI for a draft of the “Why it matters” section of T&S Insider, because I'm too close to the content or just run out of time. I used to sometimes just leave it to Ben to do (sorry Ben!). At least now I’ll always present him with something halfway decent to start from.
• Writing for work - At my regular day job, I’m responsible for writing blog articles, LinkedIn content, product briefs, customer FAQs, strategy documents, and policy drafts (among other things). I do a lot of writing at work. I find AI to be particularly helpful at these tasks, because I’m less concerned about my personal voice/tone shining through, and the outputs are much more straightforward. I’ve created custom Gems with our sales decks and other basic info, and use that info as additional background context for any content planning I’m working on, which saves a lot of time. 

Outside of writing, I love using Gemini for work productivity. It’s connected to my calendar, docs, meeting notes, and email so, every morning, I ask for a brief on what my most important tasks are. I also use it to keep me on track for quarterly goals and strategic planning. Custom Gems are underutilised, in my opinion.

My philosophy for AI

Overall, I use AI as a sounding board, not a ghost writer. It’s helped me to consider other perspectives, sharpen my arguments, and be less long-winded. I work best when I can "think out loud" for a while with someone, and AI lets me do that. AI has helped me refine my writing process and speed up the amount of time from first idea to final draft.

However, the best feedback and ideas I get are always (always) from Ben and from readers like you. The fun of writing T&S Insider is that I'm creating resources and stating opinions that don't exist online yet. That can never be replicated by AI.

Also worth reading

T&S Specific Measures Survey (Trust & Safety Foundation)
Why? Take a few minutes to complete this survey. It's part of a research initiative aimed at advancing the understanding and support of mental health and wellbeing among professionals in T&S roles.

For the fellow parents reading, a trio of links on kids and phone use:

Is Jonathan Haidt right about smartphones? (tes magazine)
Why? A nuanced and thoughtful look at Haidt's claims about phone use. Also mentioned on this week's Ctrl-Alt-Speech.

This counterintuitive advice about kids and phones makes a lot of sense (Vox)
Why? This piece includes the quote "The case for giving one to your 3-year-old is not as outlandish as it sounds." Not sure I'll be giving my kid a phone yet, but I agree that totally shielding your kid from technology while also using it all the time is not sending the healthy signal many think it is.

Raising Digital Natives
Why? A practical and sensible toolkit for parents on how to evaluate technology and talk to kids about it in an age-appropriate way.