Etiquette, Accuracy and Avoiding the AI Voice

You now have the practical skills to draft, tone-match, personalise and edit almost any professional email with AI help. This final lesson covers the habits that hold everything else together: the etiquette questions people worry about but rarely discuss openly, the accuracy checks that protect you from a genuinely damaging mistake, and the specific tells of “AI voice” that, once you can spot them, you will never miss again in your own drafts or anyone else’s.

Do you have to disclose that AI helped write it?

There is no universal rule requiring disclosure that you used AI to draft a routine business email, and for the vast majority of everyday professional correspondence, status updates, meeting requests, follow-ups, disclosure is neither expected nor useful, in the same way nobody expects disclosure that you used spellcheck or a thesaurus. The relevant ethical line is not whether AI was involved, but whether the content is accurate, whether it represents your genuine views and commitments, and whether you have taken responsibility for what it says by reading it properly before sending. Where disclosure becomes more relevant is in specific contexts: academic or journalistic writing with explicit AI use policies, formal statements where authorship itself is the point, or any situation where your organisation has a specific policy requiring it. Know your own employer’s policy on AI tool use for external communication, since these policies vary and are becoming more common.

The accuracy rules that matter most

Across every lesson in this course, one theme recurs because it is the single biggest real risk of AI-assisted email: language models can produce fluent, confident, plausible-sounding text that is factually wrong, a fabricated statistic, an invented client name, a misremembered date, a policy that does not actually exist. This tendency, sometimes called hallucination, is not a rare glitch, it is a structural feature of how these models generate text, predicting plausible next words rather than checking a database of verified facts unless specifically connected to one. Treat this as a permanent operating assumption, not a problem that will disappear as the tools improve.

Spotting and removing the AI voice

Readers are increasingly good at detecting when an email was clearly AI-generated and barely edited, and that perception, fair or not, can undercut trust even when the content itself is accurate. A handful of specific patterns give it away most often. Overuse of the word “delve,” along with other words models favour disproportionately such as “boast,” “tapestry,” “realm” and “leverage” used as a verb, reads as noticeably stilted in everyday business email. Triplet phrasing, listing exactly three adjectives or examples in a row, “innovative, efficient, and scalable,” appears far more often in AI output than in natural human writing and becomes noticeable once you know to look for it. Overly balanced, hedge-everything structure, always presenting two equally weighted sides before landing softly in the middle, reads as evasive in a business context where the reader wants your actual view. Excessive use of em dashes to link clauses is another common tell, and one this course itself deliberately avoids throughout.

The fix is not to avoid AI, it is to edit what it gives you. Read every draft with this checklist in mind, and when you spot one of these patterns, cut it directly or ask the tool to revise: “Rewrite this without triplet lists, without the word delve or leverage, and without over-balancing both sides of the point, just say what you actually think.” Over time, both your prompts and your ear for editing will make these patterns rarer in your first drafts.

The habit that matters more than any single technique

If this course leaves you with one durable habit, it should be this: never send an AI-drafted email you have not personally read in full, start to finish, immediately before sending. Not skimmed, not trusted because the first two sentences looked right, read properly, checking facts, checking tone, checking that it sounds like something you would actually say. Every technique in this course, structured prompting, tone control, personalisation, editing passes, template libraries, exists to make that final read faster and more likely to require only small fixes. None of it exists to replace that final read altogether, because you, not the AI, are accountable for what lands in someone else’s inbox with your name on it.

Where to go from here

You now have a complete, practical system: how to choose your tools, structure a prompt, get the tone right, handle everything from a routine update to a difficult apology, personalise at scale, edit ruthlessly, and build a template library that keeps getting better. The remaining step is simply repetition. Use these techniques on real emails this week, notice what works for your specific role and audience, and adjust your prompts and templates accordingly. Within a month of consistent practice, drafting a strong, well-toned professional email with AI assistance will feel less like a skill you are learning and more like a habit you have simply always had.