Why we're telling you when it's AI, even when we don't have to

Somewhere in Brussels, a piece of legislation is being finalised that will eventually tell every company using AI in Europe how much they need to explain themselves to the people affected by it. The EU AI Act sets out rules on transparency, on human oversight, on knowing when a machine has had a hand in a decision that matters to you. Most of it doesn't formally apply to a fit-scoring tool used by a nineteen year old in Manchester. But we read it anyway and found ourselves asking a more useful question than "what are we obliged to do", which I'm guessing is the question being asked more broadly. Instead, we asked what we'd want to know if we were the one applying for the job.

That question is the whole reason Hike exists. We built this company because the job market lies to people by omission. Rejections with no reason. Job descriptions that don't match the actual role. CVs that vanish into applicant tracking systems nobody ever explains. Honesty was never a marketing line for us, it's the constraint the entire product sits inside. So when it came to AI, the same logic applied. If we're going to use it to help you understand your own CV, your own fit for a role, your own gaps and strengths, you should know exactly where the machine's fingerprints are and where they aren't.

That's what our new AI transparency page is for. We've gone through every feature on the platform, Job Fit analysis, Career Health Check, Targeted CVs, Bias detection, the Career Sherpa, Work Diary Insights, and split out what the AI is actually doing from what's calculated by fixed, deterministic code with no model involved at all. Your fit score, for instance, is never invented by the AI on the fly. It's produced by a scoring formula we built and maintain, applied to structured output the AI generates from reading your CV and the job description. The AI reads and writes. The number is ours and it's the same number every time given the same inputs.

We've also been straightforward about where AI gets it wrong.

It can be confidently wrong, stating something inaccurate with total fluency. It can carry biases from its training data, which is exactly why our bias detection work is grounded in research rather than the model's own instincts. It doesn't know your target employer, their culture, or their hiring quirks, unless you've told Hike yourself. And if you're using a Hike-optimised CV, you're still the author of record. And we're super-clear: read the whole thing before you submit it anywhere.

None of this is the most exciting thing we've shipped this year. There's no flourish to a page full of tables explaining what's deterministic and what isn't. But we think it might be one of the more important things we've built because it's a public promise that's genuinely hard to walk back from. Once you've told people exactly what your AI does and doesn't do, you've committed to it.

We're building a product for people who've had enough of systems that don't explain themselves so being transparent about our own AI isn't a compliance exercise for us. It's the same honesty-first promise we make about everything else on the platform, just pointed at ourselves this time.

If you want the full detail, including your rights under EU and UK AI governance frameworks and how to flag anything you think we've got wrong, it's all on the AI transparency page. And if you spot something that doesn't sit right, tell us. That's rather the point of the whole exercise.