You click Help me write in Gmail, type a one-liner like “decline politely, suggest next month,” and Gemini hands back a generic three-paragraph corporate email that opens with “I hope this message finds you well” and never references the thread you are replying to. Or the tone is off: too formal for a coworker you Slack daily, too breezy for a vendor. You polish it by hand, which defeats the point.
Fastest fix: open the email and click Reply first, then invoke Help me write from inside the reply (not Compose), give it a three-line prompt (action, reason, offer), generate once, then tap the Friendly or Shorten refine button. If drafts still feel generic on a personal account, turn on smart features in Settings > See all settings > General > Smart features so Gemini can actually read your thread and past mail.
As of June 2026 the model behind Help me write is Gemini 3 (Google’s Gemini 3.1 Pro family), the same model that powers the standalone Gemini app. Google moved most of these tools inline in Gmail during 2026 rather than burying them in a separate side panel, and Help me write is now free to everyone, not just paid plans. So when the email comes out wrong, the bottleneck is almost never raw model capability. It is context, settings, or prompt scaffolding. This article walks through each cause and the exact fix.
First, check eligibility (5 seconds)
Help me write is available worldwide on an eligible Google Workspace or Google AI plan. On a personal @gmail.com account it is US-only, in supported languages, as of June 2026. If you are outside the US on a personal account, the button may be absent or limited. That is a rollout boundary, not a bug.
The separate Proofread feature (advanced grammar, tone, and style check before you send) requires Google AI Pro ($19.99/mo) or AI Ultra ($99.99/mo). Help me write itself does not.
Common causes
Ordered by hit rate, highest first.
1. Gemini does not have the thread
The single biggest reason. Help me write inside a new Compose window has no thread to read. Inside a Reply it can attach the thread, but only if you opened the reply first and then invoked Help me write. If you opened Compose and pasted the original email into the prompt as background, Gemini still treats it as a blank draft.
How to spot it: drafts that ignore specific names, dates, or numbers from the thread. If Gemini writes “regarding our previous conversation” instead of “regarding the Q3 budget you sent Tuesday,” it cannot see the thread.
2. Smart features are off
This is the cause most people miss. The toggle at Settings (gear) > See all settings > General > Smart features (“Turn on smart features in Gmail, Chat, and Meet”) is what lets Gemini read the structure and content of your thread and lean on your past mail. In the EEA, Japan, Switzerland, and the UK this is OFF by default (other regions default ON). With it off, Help me write falls back to a context-free, neutral-formal template every time.
How to spot it: every draft is generic no matter what you do, and the thread is never referenced. Check the toggle before blaming the prompt.
3. Tone preset defaults to neutral-formal
The first draft has no tone overlay applied; it defaults to a neutral, slightly formal B2B voice. The refine buttons that change tone, as of June 2026, are Formalize, Friendly, and Shorten on desktop (with Polish, Elaborate, and a free-text “Describe your change” box also available depending on rollout). For a long-time coworker, the un-refined default reads cold.
How to spot it: the opening line is “I hope this email finds you well” or “Thank you for reaching out.” That is the default tone signature.
4. Prompt is too short and too vague
“Decline politely” gives the model no constraints, so it fills the gap with boilerplate. The Help me write UX makes short prompts look fine, but the underlying model needs the same scaffolding any LLM does: who you are to the recipient, what specifically you are declining, what you want to offer instead.
How to spot it: your prompt is under 10 words and the output is several paragraphs long. The model is improvising most of the content.
5. Drafts are not in English (style matching)
Google’s own requirement: to get Help me write to match your tone and style, your drafts must be in English as of June 2026. In other languages you still get a draft, but the personalize-to-you pass does not run, so output reverts to a neutral template.
How to spot it: non-English drafts feel flat and interchangeable even when your English drafts in the same account sound better.
6. Personal account with no style signal
On a personal account (or a brand-new Workspace seat), there is little or no sent-mail history for Gemini to anchor on, so it falls back to the default voice. Workspace accounts with smart features on and a real sent-mail history get more personalized output.
How to spot it: drafts feel generic regardless of prompt. Check the Sent folder. Only a handful of substantive sent emails means the style anchor is weak. In-prompt anchoring (Step 5 below) fixes this regardless of account type.
7. Wrong language detection on mixed threads
You write the prompt in English but the thread is in Chinese, Japanese, or Spanish. The model may detect the prompt language and draft in English, or detect the thread and draft in the thread language, and it can guess wrong about which you wanted.
How to spot it: output language is not what you expected. Re-prompt explicitly: “reply in Mandarin” or “reply in English.”
8. Sensitive-topic redirect
If the thread mentions legal, medical, or financial advice, or HR-coded keywords (termination, harassment, layoff), Gemini falls back to extremely cautious neutral-corporate phrasing regardless of your prompt. This is a safety pattern, not a tone bug.
How to spot it: drafts are noticeably stiffer for one specific thread but fine for others. Check the thread for sensitive triggers.
Which bucket are you in?
| Symptom | Most likely cause | Go to |
|---|---|---|
| Every draft generic, thread never referenced | Smart features off, or composed in a new window | Step 0, Step 1 |
| Tone too formal for the recipient | No tone refine applied | Step 3 |
| Output ignores who/what/when of the request | Prompt too thin | Step 2 |
| Generic only in non-English drafts | English-only style matching | Step 6 |
| One specific thread is stiff, others fine | Sensitive-topic redirect | Step 7 |
| Button missing entirely | Eligibility / region / admin restriction | ”When this is not on you” |
Shortest path to fix
Step 0: Turn on smart features
Go to Settings (gear) > See all settings > General, scroll to Smart features, and check Turn on smart features in Gmail, Chat, and Meet. If you are on Workspace, there may be a second control for smart features in other Google products. Save. This is the prerequisite for thread-aware drafting; do this before anything else if drafts are uniformly generic.
Step 1: Open the reply first, then Help me write
Click Reply on the thread, then invoke Help me write from inside the reply window. That attaches thread context. Do not click Compose and try to paraphrase the thread into the prompt.
Step 2: Give a three-part prompt
Replace “decline politely” with three short clauses, each one constraint:
Decline the Tuesday meeting invite from Jane.
Reason: I am OOO that week.
Offer: any time the following Monday or Tuesday.The model now has enough to write a draft that references the thread and gives a concrete next step.
Step 3: Refine after the first draft, not before
Generate the draft, then use the refine buttons: Friendly to warm the tone, Shorten to cut length, Formalize to tighten up. Friendly applies a tone overlay the initial draft does not have. One or two refine taps usually lands the tone within a click of correct. If your build shows a free-text “Describe your change” box, you can also type “drop the greeting line, sound like a quick note to a teammate.”
Step 4: Add a tone anchor in the prompt itself
Append an explicit tone instruction, for example: tone: like a quick note to a coworker, no greeting line. The model honors explicit tone instructions far more reliably than it infers tone from context.
Step 5: For personal accounts, paste a voice sample
If you do not have Workspace and the default tone is consistently wrong, paste two lines from one of your own past replies at the top of the prompt:
Match this voice: "yeah that works, lmk what time. I can do Tues after 3."
Now draft a reply that declines Wednesday and offers Thursday.This gives the model an in-prompt style anchor regardless of account type or sent-mail history.
Step 6: Draft in English for best tone matching
If your account is set up so style matching applies (English drafts), Help me write personalizes more. For other languages, expect a more neutral baseline and lean harder on Step 4 and Step 5.
Step 7: Switch language explicitly on mixed threads
State the output language directly: Reply in English or 用中文回复. Do not rely on auto-detection for mixed-language threads.
Step 8: For sensitive threads, draft manually
If the thread touches legal, medical, or HR topics, expect Gemini to keep the corporate tone no matter what you prompt. Either write those replies by hand, or strip the sensitive keywords from your prompt and add them back in a final manual pass.
How to confirm it is fixed
A correctly working draft will (1) name something specific from the thread (a date, a name, a number), (2) carry the tone you asked for instead of “I hope this email finds you well,” and (3) stop at the length you wanted after one Shorten tap. If you click the Sources control under the draft, you should see the thread (and, with personalization on, related mail) listed as what Gemini read. No sources listed usually means it had no context, which sends you back to Step 0 and Step 1.
When this is not on you
Google ships A/B variants of the Help me write prompt template and refresh cycles for the Gemini 3 family. Some weeks the default tone is noticeably warmer; other weeks it tightens. If quality regressed suddenly across all your threads with no change on your end, you may be on a new variant. Wait a week or two and re-evaluate.
Workspace admins can also restrict Gemini features in the Admin Console, including the smart-features controls. If Help me write is missing entirely on a managed account, ask your admin. On a personal account outside the US, the absence is the rollout boundary noted above, not an admin setting.
Easy to misdiagnose as
Assuming the model is “just bad at email.” It is the same Gemini 3 that drafts fine in the standalone app; the constraint is missing context and settings, not model capability. Moving the same request to gemini.google.com with the thread pasted in usually produces a noticeably better result, which confirms the issue is on the Gmail/context side, not the model side.
Also easy to confuse with Smart Compose (the gray inline autocomplete) and the newer Suggested Replies (one-click responses under a thread). Those are sentence-level or whole-reply suggestions using faster, lighter logic, and they share the same smart-features toggle. They are separate from Help me write.
Prevention
- Turn on smart features once, then leave it on (region permitting).
- Always click Reply first, then Help me write. Never the other way around.
- Use three-clause prompts: action, reason, offer.
- Append an explicit tone anchor for any draft going to someone you know well.
- Keep a 2-line “voice sample” snippet ready to paste for casual replies on personal accounts.
- For language-mixed threads, state the output language in the prompt.
- Accept that legal, HR, and medical threads will draft stiffly, and plan to hand-edit those.
FAQ
- Which model writes my Gmail drafts now? As of June 2026, Help me write runs on Gemini 3 (Google’s Gemini 3.1 Pro family), the same model as the standalone Gemini app. The difference in output quality comes from context and settings, not the model.
- Why does the standalone Gemini app sometimes write better email than Gmail? When smart features are off or you composed in a fresh window, Gmail has no thread context to feed the model. Paste the thread into the app and it has everything; that is the gap, not model quality.
- Is Help me write free? Yes, as of June 2026 it is free to all Gmail users (personal accounts US-only). The separate Proofread tone-and-style checker requires Google AI Pro or Ultra.
- Why is the button missing? On a personal account outside the US it has not rolled out. On a managed Workspace account an admin may have disabled Gemini or smart features in the Admin Console.
- Will Gemini learn my voice over time in Gmail? With smart features on and an English sent-mail history, drafts pick up some of your style, but it is not personalized fine-tuning. For close matching, paste an in-prompt voice sample.
- Does turning on smart features share my email content? It lets Gmail and Gemini use your mail to personalize drafts and suggestions. Google states this info is used to improve your drafts in Help me write. You can turn the toggle off any time in the same settings panel.
Related
- Gemini Extensions (Workspace, Maps, YouTube) Not Triggering
- Gemini Workspace Integration Issues
- Gemini Workspace Not Syncing
- Gemini Replies in the Wrong Language
- Gemini App and Web Give Different Answers
- Gemini Doc Summary Is Weak
- Gemini Google Account Permission Errors
- Gemini Gems Not Saving
- Gemini Saved Info Not Persisting
- Gemini Context Too Short
Tags: #Gemini #Troubleshooting #google-ai #Gmail #Workspace #help-me-write