Summarize with AI
Good HR tech website design answers the fairness question before the buyer has to ask it. An HR leader looking at an AI hiring tool wants to know three things early: how you test it for bias, who checks its decisions, and what the candidate is told. If the homepage only promises speed, she reads you as the risky vendor, however good the product is.
This matters more for an AI startup than for an old applicant tracking system. The person who buys your tool is the person whose name sits next to it when a rejected candidate complains. She is not shopping for the fastest screen. She is shopping for one she can defend. The site that shows it can be defended reads as the serious company in the category, and the one that hides the question reads as the one that has not thought about it.
Hire the best, 10x faster, and the word bias appears nowhere
Picture an AI screening tool, used here as an example. It reads every application for a role, scores each one against the job description, and hands the recruiter a ranked shortlist of twenty. It is sold to companies with 200 to 2,000 staff that hire in bulk, like support teams and warehouses.
The homepage hero reads Hire the best, 10x faster. Under it sits a stock photo of a smiling candidate on a video call, and a purple Book a demo button. Then a logo strip of five names. Then three feature cards: Smart screening, Instant shortlists, Automated outreach. The nav has Product, Pricing, Customers and Book a demo. The footer has a Privacy link and a Terms link.
Now picture the head of talent acquisition at a 900-person retailer with stores in New York. She opens the page in one tab of six. She presses Command F and types bias. Zero results. She types audit. Zero. She types fair. One hit, inside the word fairly, in a testimonial. She opens the Privacy link, scrolls past the cookie table, finds nothing about how the scores are made, and closes the tab. In her shortlist spreadsheet the column for this vendor reads, in the kind of shorthand buyers use, no audit info.
Nobody tells the founder this happened. There is no form fill, no bounce reason, no email. The page did exactly what it was built to do, which was to sell speed, to a reader who came to check for risk.
Why speed is the wrong first promise for HR buyers
Most website advice says lead with the outcome and keep compliance on a page in the footer. For most software that is right. For AI that ranks people, it gets the order backwards. Here the fairness proof is part of the outcome, because a shortlist she cannot defend is not a shortlist she can use.
Think about what the HR buyer is answerable for. Her CEO asks about time to hire. Her general counsel asks whether the tool could screen out a protected group. A candidate who was rejected can ask why. Speed helps with the first conversation. It is a liability in the other two, because a faster biased screen does more harm per week. So the claim 10x faster, standing alone, makes a careful reader slightly more nervous, not less.
This is the same trust problem a security buyer brings to a product page, and the cybersecurity website guide covers that reader. The HR version is sharper in one way. The person at risk is not the buyer's company alone. It is every candidate the model reads, and they never chose your product.
What the founder sees from his side of the call
The founder does not see the closed tab. He sees what reaches his calendar. Demos come mostly from outbound. The first call goes well, because he shows the shortlist and the buyer nods. The second call gets moved twice, and when it happens there is a new person on it from legal or people operations. Her first question is the one his homepage never answered.
So he answers it by email. He writes the same three paragraphs about how the model was tested, pastes them into four threads in one month, and attaches a PDF that was made the week before a pilot. He tells the team that HR is a slow market. He adds a line to the board update saying sales cycles are long in the category. Both statements are true, and both skip the part he controls.
The cost is time, and time is the scarce thing at seed. The median gap from seed to Series A is around 616 days across all sectors, and roughly 70 percent of seed companies never raise one. For AI startups the Series A bar sits around 3.5 million in ARR, up from roughly one million three years earlier (Carta, Q1 2026). Every deal that waits a month for a fairness answer is a month of that runway, spent on a question a page could have answered in one scroll.
The fairness section, block by block
Put it on the homepage as the second screen, right after the hero and before the feature cards. Give it a plain heading, something like How we keep screening fair, and four blocks under it. Each block is a short claim and a link to the proof.
- What the model reads and what it does not. For the example tool: it scores skills and experience against the job description, and it does not see names, photos, dates of birth or home addresses. Write only what your product actually does, and say it in words a recruiter would repeat to a candidate.
- Who tested it, when, and where the results live. Name the independent auditor, the date of the last audit and the roles it covered, and link the published summary. If you have not had an audit yet, say what testing you run today and when the first audit is booked. An honest gap reads better than silence.
- Where a human decides. Say exactly which steps a recruiter must approve. For example, no candidate is rejected without a recruiter clicking Approve, and every score shows the three reasons behind it.
- What the candidate sees. Show the real notice a candidate receives, the line on the application screen that says an automated tool is used, and how a candidate can ask for a different process. A screenshot of that screen does more than a paragraph about candidate experience.
Use real product screens for all four, with sample data. A recruiter view where each score lists its reasons, an audit summary page, the candidate notice. HR buyers have seen the smiling stock photo on every vendor site. They have rarely seen the screen where a recruiter overrides a score, and that screen is what tells them a human is still in charge.
Name the law only as far as it goes
Regulation is the part founders either skip or overclaim. Do neither. State the rules your customers face, accurately, link the source, and say what your product does to help. Then have your own counsel read the page before it goes live. A website is not legal advice and should not sound like it.
Two rules are worth knowing by name. In New York City, Local Law 144 says an employer may not use an automated employment decision tool unless it has had a bias audit within one year of use. The employer has to post a summary of the audit results and give candidates notice before the tool is used. The city began enforcing it on July 5, 2023. Note who carries the duty: the employer or employment agency, not the vendor. So the honest line on your site is that you supply what customers need to comply, not that you make them compliant.
In Europe, the AI Act lists AI systems intended for the recruitment or selection of people among its high-risk uses, including screening applications and evaluating candidates. If you sell to European employers, a short Compliance by region page that names this, and says what documentation you can share, answers the question their legal team will ask. Keep dates and duties to what you can link, because these rules are still being phased in and a wrong date on your homepage is worse than none.
Five changes to make to the site this week
None of these needs a redesign. A founder and one engineer can do most of them in a few days.
First, run the buyer's search on your own homepage. Press Command F and type bias, audit, fair, human, candidate and notice. Write down every zero. Each zero is a question a buyer asked and left with no answer.
Second, copy your best email answer onto the page. The three paragraphs you keep pasting into sales threads are the first draft of the fairness section. Cut them to four blocks and put them second on the homepage.
Third, swap the stock photo for a product screen. Use the recruiter view that shows why a candidate scored the way they did. If that screen does not exist in the product yet, that is a product gap, not a website one, and it is worth knowing.
Fourth, give fairness its own page, linked from the main nav, not only the footer. Put the audit summary, the testing method in plain words, the candidate notice and a contact for questions on it. Call it Fairness or Trust, not Legal.
Fifth, rewrite the hero so speed and defensibility sit together. For the example tool, Shortlists your recruiters can explain, in a day instead of a week says more to an HR buyer than Hire the best, 10x faster. Keep the speed. Attach it to the thing she is judged on.
Two readers, one page, and where to go next
An HR tech site talks to two rooms at once: the HR leader deciding whether the tool works, and counsel deciding whether it is safe to use. The same was true for the Mi-VAD heart pump website, where investors were deciding whether the device was real and clinicians whether it was safe, and both formed a view of the company before they reached the science. The broader structure of a site that meets a buyer's main objection where it arises is in what makes a good SaaS website. If you would rather hand the work to a studio, the shortlist of website design agencies for AI HR tech startups compares teams on public criteria.
A fairness section only works if the buyer finds it in seconds and believes what she reads, which makes it a design job as much as a legal one. Studio Maydit designs that kind of page for AI founders in the US, UK and Europe, building in Framer, Webflow or custom code, and the work can continue into the product screens your recruiters and candidates use. A fixed-scope project takes three to four weeks. If your second calls keep opening with the question your homepage skips, book a 45-minute call with Studio Maydit.
