Using Gemini, Claude and Copilot on your CV: what their terms say
All three chatbots can draft or polish a CV, and each vendor's own terms warn that answers can be wrong. They differ on your data: whether chats train models, how long they are kept and whether people review them.1
Written 13 October 2026. CitedCV. How we write our guides
Before you paste: make a CV copy for the chatbot
Whichever assistant you choose, paste a version of your CV that still lets it improve the wording but carries nothing a stranger could use. Google's Gemini privacy hub asks you not to enter confidential information you would not want a reviewer to see. Microsoft's Copilot terms ask you to review responses before sharing them so you do not expose personal information.
| CV detail | Paste it? | Why |
|---|---|---|
| Your name | Replace with [NAME] | Better wording does not depend on it |
| Email, phone, home address | Leave out | Add them back in the final file |
| Date of birth, ID or passport number | Never | No edit to a CV needs them |
| Referees' names and contact details | Leave out | They are other people's details |
| Job titles, employers, dates | Keep | The draft needs them, and they are what you will check afterwards |
| Client names covered by a confidentiality clause | Describe instead, such as a national retail chain | Your contract decides this, not the chatbot |
This table is our advice, not a rule from any of the three vendors.
The settings that decide what happens to your CV
Read on 13 October 2026 from each vendor's own help and privacy pages, for personal accounts: Gemini with a personal Google Account, Claude on the Free, Pro and Max plans, Microsoft Copilot with a personal Microsoft account.
| Question | Gemini | Claude | Copilot |
|---|---|---|---|
| Are chats used for training? | Yes while Keep Activity is on: Google says it uses activity to improve its services, including training generative AI models | Only if you allow it, if a chat is flagged for safety review, or if you send feedback | Not in the app released on 18 August 2026: prompts, responses and file contents are not used to train foundation models |
| Where to switch it off | Keep Activity, managed in Gemini Apps Activity | Settings, Privacy, Help improve our AI models | The new-app pages describe no training switch; the older app had an opt-out |
| How long chats are kept | Auto-deleted after 18 months by default (3 or 36 months, or never); 72 hours when Keep Activity is off | Deleted chats leave back-end storage within 30 days; up to 5 years, de-identified, if you allow training | The older app kept conversations 18 months by default; the new-app pages we read give no period |
| Do people read chats? | A subset is reviewed; reviewed chats are kept up to 3 years and survive deleting your activity | Chats flagged by safety systems: inputs and outputs kept up to 2 years | Not stated on the pages we read |
| Private mode | Temporary chats: not used for training, kept 72 hours | Incognito chats: not used for training, kept 30 days | Not covered on the pages we read |
| Thumbs up or down | Reviewed feedback and the related chat are kept up to 3 years | The whole conversation is stored up to 5 years and may be used for training | Optional feedback is not used to train the foundation models |
| Work or school account | Different terms, in the Workspace privacy hub | Commercial products are not used for training by default | Not covered by these pages; Microsoft has a separate page for Entra accounts |
Copilot has two sets of pages at the moment
Microsoft released a new Copilot app on 18 August 2026 and now keeps two sets of privacy pages. The older Privacy FAQ, which applies only to the older app, says Microsoft uses Copilot conversations, including uploaded images and files, for AI training unless you opt out or belong to an excluded group, and that it removes identifying information first. The newer pages say prompts, responses and file contents in the new app are not used to train foundation models. Check which app you are using before relying on either.
Gemini keeps a copy even with history off
Switching Keep Activity off does not make a chat disappear at once. Google keeps such chats with your account for 72 hours so Gemini can respond and for safety. Chats that a reviewer has already seen are disconnected from your account and kept for up to three years, even if you later delete your activity.
What each vendor says about wrong answers
None of the three promises accuracy, and each puts the warning in its own words:
- Google: LLM experiences, Gemini included, can hallucinate and present inaccurate information as factual, and Gemini may produce inaccurate information, including about people.
- Anthropic: Claude's outputs may contain material inaccuracies even when they appear accurate because of their level of detail or specificity, and you should not rely on them without confirming them yourself.
- Microsoft: Copilot can make mistakes and may give responses that seem convincing but are incomplete or inaccurate. Its transparency note adds that it could fabricate content that sounds reasonable but is factually inaccurate.
Why the most specific line deserves the most doubt
Anthropic's wording names the trap on a CV. A bullet such as cut onboarding time by 32% for a team of 14 reads as credible precisely because it is detailed, and detail is what a model supplies when it fills a gap you left. The guide to why AI CV builders hallucinate explains the mechanism; the checks below follow from the vendors' own warnings.
Four checks that come from the vendors' own warnings
- Detail you did not supply. Put your original next to the draft and mark every number, tool, title and date that appears only in the draft. Confirm each one from a record or delete it.
- Web content mixed in. Microsoft's transparency note says Copilot grounds information-seeking conversations in web search results. When polishing a CV, our suggestion is to turn off Web search in the new app (Settings, Personalization) so text about your employer, or about someone with your name, does not drift in.
- Wording other people also get. Copilot's terms say responses may not be unique to you and that others may receive the same or similar ones. A summary you did not rewrite may be a summary other applicants send as well.
- Disclosure. Copilot's code of conduct asks you not to mislead others about whether you used Copilot to create your content. If an application form asks whether you used AI, answer it plainly.
The first check on an example
Invented example to show the method; it is not output from any of the three tools.
| What you gave | What the draft says | What to do |
|---|---|---|
| Ran weekly stock counts at a warehouse | Led weekly stock counts for a team of 12, cutting discrepancies by 18% | Delete the team size and the percentage unless you hold a record of both |
| Used Excel for reports | Built Power BI dashboards and Excel reports | Remove Power BI if you have not used it |
| Logistics assistant, 2022 to 2024 | Logistics coordinator, 2021 to 2024 | Restore your real title and start year |
Sources, read on 13 October 2026
| Page | Date shown | What we took from it |
|---|---|---|
| Google: Gemini Apps Privacy Hub | Last updated 13 October 2026 | Keep Activity, 72 hours, auto-delete, reviewers, 3 years, accuracy |
| Anthropic: Is my data used for model training? (consumer) | 16 March 2026 | When chats are used for training; incognito; feedback |
| Anthropic: How long do you store my data? | 1 July 2026 | 30 days after deletion, 5 years, 2 years for flagged chats |
| Anthropic: model improvement privacy settings | 3 August 2026 | The settings path |
| Anthropic: Consumer Terms of Service | Effective 8 October 2025 | Reliance on outputs |
| Microsoft: Copilot activity history (new app) | New app available 18 August 2026 | No training on prompts, responses or file contents |
| Microsoft: Privacy FAQ for Microsoft Copilot (older app) | No date shown | Training unless you opt out, 18 months |
| Microsoft: Copilot Terms of Use | Effective 18 August 2026 | Mistakes, responses not unique, disclosure |
| Microsoft: Copilot transparency note | No date shown | Web grounding and fabricated content |
When not to paste your CV into a consumer chatbot
- Your employer's policy forbids putting work information into AI tools, and your CV describes current internal projects.
- You are signed in with a work or school account. Its data terms are your organisation's, not the consumer terms in the table.
- You have no records of your own. Without contracts, payslips or notes on titles, dates and numbers, the first check above cannot be done.
Where to go next
For a prompting routine with one assistant, from fact sheet to paste test, see the guide to writing a resume with ChatGPT; its steps work the same way in Gemini, Claude and Copilot. The AI CV builder guide sets chat assistants against grounded builders.
If you would rather not check every line by hand
CitedCV takes a different route: it writes only from facts you give it, each line carries the ids of the facts it came from, and lines whose numbers or names cannot be verified against those facts are dropped. Text you upload is scanned for prompt-injection patterns, and no fact is taken from a flagged line.
Questions people ask
Is it safe to put my CV into Gemini?
It is safer with Keep Activity off or in a temporary chat. Google says a subset of chats is read by human reviewers and kept up to three years, and asks you not to enter confidential information. Remove contact details and ID numbers first.
Does Claude train on my CV?
Only if you allow it under Settings, Privacy, if the chat is flagged for safety review, or if you send thumbs-up or thumbs-down feedback. Incognito chats are not used for training, according to Anthropic's privacy center.
Does Microsoft Copilot use my resume to train its AI?
Not in the new Copilot app, according to Microsoft's pages for it: prompts, responses and file contents are not used to train foundation models. The older app's FAQ described training unless you opted out.
Which is better for a CV: Gemini, Claude or Copilot?
None of them claims to be accurate; all three warn that answers can be wrong. Choose on data settings such as training, retention and human review, and run the same checks on whichever draft you get.
Can Gemini, Claude or Copilot make up details on my CV?
Yes, each vendor says its tool can produce wrong information. Anthropic's terms add that inaccurate output can look accurate because it is detailed, which is how an invented metric on a CV looks.
Should I delete the chat after editing my CV?
It helps but does not erase everything. Claude removes deleted chats from back-end storage within 30 days, while Google keeps chats a reviewer has already seen for up to three years after you delete your activity.
Do I have to tell an employer I used Copilot?
Answer honestly when an application asks. Copilot's terms ask users not to mislead others about whether they used Copilot to create their content; beyond that, follow the employer's own policy.
Is Copilot with a work account different?
Yes. Microsoft's consumer Copilot privacy pages do not apply when you sign in with a work or school (Entra) account; your organisation's terms and a separate Microsoft page cover that use.
Sources
- Google: Gemini Apps Privacy Hubsupport.google.com/gemini/answer/13594961?hl=en
- Anthropic Privacy Center: Is my data used for model training?privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training
- Anthropic Privacy Center: How long do you store my data?privacy.claude.com/en/articles/10023548-how-long-do-you-store-my-data
- Anthropic Privacy Center: How do I change my model improvement privacy settings?privacy.claude.com/en/articles/12109829-how-do-i-change-my-model-improvement-privacy-settings
- Claude Help Center: Use incognito chatssupport.claude.com/en/articles/12260368-use-incognito-chats
- Anthropic: Consumer Terms of Servicewww.anthropic.com/legal/consumer-terms
- Microsoft: Copilot for individuals, your activity historysupport.microsoft.com/en-us/privacy/microsoft-copilot/activity-history
- Microsoft: Privacy FAQ for Microsoft Copilot (older app)support.microsoft.com/en-us/microsoft-copilot/privacy-faq-for-microsoft-copilot
- Microsoft: Copilot Terms of Usewww.microsoft.com/en-us/microsoft-copilot/for-individuals/termsofuse
- Microsoft: Copilot transparency notesupport.microsoft.com/en-us/privacy/microsoft-copilot/transparency-note