How our AI tools work
The AI email writer, cover letter generator, paraphrasing tool, text summarizer, interview answer generator and counter-offer email generator all do the same thing: they take what you typed, wrap it in an instruction we wrote, send it to a large language model run by a third party, and show you the reply unchanged. No human reviews it. No fact-checking step exists between the model and your screen.
Which provider processes your text
Text you submit leaves our server and is processed by one of these companies, over an encrypted connection, under their terms and their privacy policy:
- Free and Pro runs using the site's key go to Groq, which is our current default provider. Pro runs may be routed to a larger model from the same or another configured provider.
- Runs using your own API key go to whichever provider that key belongs to. You can currently bring a key for Groq, OpenAI, Google (Gemini API), Anthropic, OpenRouter. In that case the request is made against your account with that provider, you pay for the tokens, and that provider's terms — including whatever it says about training on API data — govern the request rather than ours.
We send the text and the tool's instructions. We do not send your name, email address or account identifier with it. We do not store your input or the output unless your plan includes History and you have switched it on, and we never use either to train a model. The full data picture is in the privacy policy. If you would rather no third party saw a particular document at all, use the browser tools — they never upload anything — or do not paste it.
Providers change. If we switch the default provider or model we will update this page; the provider actually used for each of your runs is recorded in your account's usage log.
What the models get wrong
- Invented details. This is the big one. Models add achievements, employers, dates, metrics, degrees and certifications you never mentioned, and they write them in the same confident tone as the true parts. Read every sentence and delete anything you cannot personally stand behind.
- Invented sources. Ask for a citation, a statistic or a company fact and you may get a plausible-looking one that does not exist. Verify before you quote.
- Generic phrasing. "I am excited to apply for this opportunity at your esteemed organization" is what the training data is full of. Recruiters read hundreds of these a week. The draft is the skeleton; your specifics are what make it worth reading.
- Wrong regional conventions. Spelling (organize/organise), date order (03/04 means different months in the US and the UK), salutations, salary formats and CV-versus-resume norms often come out American by default. Fix them for the country you are applying to.
- Summaries that drop the load-bearing sentence. Summarizers compress by discarding, and they do not know which caveat matters to you. Re-read the original before you are tested or interviewed on it.
- Paraphrasing that changes the meaning. Rewording a technical, legal or medical passage can quietly invert it. Compare against the source.
- Stale knowledge. Models have a training cutoff and no live access to a company, a job posting, or this year's rules.
- Bias. Output can reflect the biases in its training data, including in how it describes people, roles and industries.
Academic and employer rules are your responsibility
Universities in the United States, United Kingdom, Canada and Australia have their own rules on AI use in assessed work, and they differ sharply — some permit AI for brainstorming and proofreading, some require a declaration, some treat undeclared use as academic misconduct. Some employers now ask whether an application was AI-assisted, and some application portals run AI-detection tools.
Check your institution's or employer's policy before you use these tools on submitted work, and follow it. When in doubt, disclose. We cannot know your rules and we are not responsible for the consequences of breaking them. Never present fabricated experience, qualifications or references as fact — quite apart from any AI policy, that can be fraud.
Not professional advice
Nothing produced by these tools, and nothing in our calculators, is legal, immigration, employment, tax, financial, medical or career advice, and using them creates no professional relationship. Notice periods, redundancy rights, visa implications and tax treatment depend on your contract and your jurisdiction. Where the stakes are real, check your contract and talk to a qualified professional.
Availability and accuracy are not guaranteed
The AI tools depend on third-party providers. They can be slow, rate-limited, temporarily unavailable, or changed without notice by the provider. Output for the same input is not reproducible. Do not plan a deadline around a generator being up — draft early.
How to use the output well
Treat every result as a first draft written by a capable but uninformed assistant who has never met you:
- Give it real specifics up front — actual responsibilities, numbers, and the actual job ad. Vague input produces vague output.
- Keep the structure, then replace every generic line with something only you could write.
- Delete any claim you cannot evidence in an interview.
- Fix the regional spelling, dates and salutation.
- Read it aloud once. If it does not sound like you, it will not sound like you to a recruiter either.
Report a bad output
If a tool produces something harmful, discriminatory, or dangerously wrong, please tell us — support@getjobtools.com or the contact form. We tune our prompts based on real reports. This disclaimer forms part of our Terms of Service.