Is an AI Reading My CV? What Applicant Tracking Systems Actually Do

A lot of current job-search advice is built on the idea that an artificial intelligence tool is reading your CV, scoring it, and then rejecting you for missing keywords. That story is mostly wrong, and believing it leads people to write worse CVs. Here’s what actually happens to your application, where automation genuinely exists, and how we handle it at 8Bit.

Where the Myth Came From?

You’ve probably seen the statistic: 75% of CVs are rejected by an ATS before a human ever sees them. We noticed it appears in career blogs, LinkedIn posts, TikToks, university career pages, and the marketing copy of nearly every CV-optimisation product.

It has a traceable origin, and it isn’t a study. The figure comes from a 2012 sales pitch by a company called Preptel, which sold resume-optimisation services. No methodology was ever published. Preptel ceased trading in 2013. Career consultant Christine Assaf traced the claim in an investigation (we highly recommend you checking out her work). Her work has since been cited by HiringThing and The Interview Guys, both of which independently document the same origin and note that searches of academic databases turn up no research supporting the number.

In other words: a company selling the cure invented the disease, went out of business, and the number outlived it. Each time it gets repeated by a larger outlet, it becomes harder to question, and the citation chain now runs several steps removed from a defunct startup with a product to sell.

The more recent evidence points the other way. A study by Enhancv interviewed 25 US recruiters across technology, healthcare, finance, publishing, and retail, collectively using more than ten major ATS platforms. As reported by IT Brief UK, HR Gazette, and TechR Series, 92% said their systems do not automatically reject applications based on formatting, design, missing keywords, or low match scores. The 8% who did use auto-rejection applied it only to roles with highly specific hard requirements (what we call the knockout questions).

Asked where the 75% figure came from, 68% of those recruiters pointed to social media posts shared by job seekers themselves, and another 20% attributed it to career coaches and CV-writing services recycling it to sell templates.

What an ATS Actually Does

An applicant tracking system is, at its core, a database with a workflow attached. Strip away the marketing and it does four things.

ATS receives and stores applications. 

Every CV, cover letter, and form response lands in one place instead of scattered across inboxes.

ATS parses your CV into fields. 

The system reads your document and tries to extract structured information: name, contact details, employment history, education, skills. This is the step where formatting genuinely matters, and we’ll come back to it.

ATS lets recruiters search and filter. 

If a recruiter needs someone with Unreal Engine experience, they search for it. Your CV appearing in that search depends on the words actually being in your CV. This is the grain of truth inside the keyword advice, but the mechanism is a search box, not an algorithm forming a judgment.

ATS helps tracking status. 

Who has been contacted, who is at interview stage, who was rejected and why. Mostly it’s an administrative record.

None of that involves an AI deciding whether you’re any good. The system organises information so a human can make decisions faster.

Where Automation (=not AI!) Is Real: Knockout Questions

Here’s the part that does get automated, and it’s worth understanding precisely because it’s the thing people mistake for AI.

Knockout questions are simple yes/no filters attached to an application form. A recruiter writes them, sets which answer disqualifies, and the platform applies that rule. Typical examples:

Do you currently hold the right to work in this country?

Are you already located in this region?

Are you able to work on-site at this location?

If a role requires someone already holding a work visa for a specific country and you answer no, the system will filter you out. That is a rule a person wrote, executing exactly as instructed. There’s no model, no scoring, no inference about your suitability. It’s an if->then statement.

This matters for two reasons. First, it’s the single most common way applications get automatically closed, so it’s worth answering these questions carefully and accurately rather than rushing them. Second, it’s frequently misread as evidence of AI screening (“I got rejected minutes after applying at midnight, so it was definitely AI”) when it’s nothing of the sort.

Where AI May Be Is Used in Hiring

It would be dishonest to say AI plays no part in recruitment. As in almost every industry, it’s becoming visible, so it may play a part in specific places, and it’s worth knowing which so you can calibrate rather than panic.

AI “fit scores” inside ATS platforms. Many Applicant Tracking Systems now offer a feature that ranks applicants against a job description. This is the closest thing to the myth, and it’s real. But to be fair, its adoption is the interesting part: our practice shows many companies and recruiters ignored it entirely. Some may treat the score as a rough guide while still reviewing manually. The bottom line is that even where the feature exists, most recruiters aren’t outsourcing the decision making part to it.

Automated video interview assessment. So platforms can score recorded interview responses. Worth knowing what this does and doesn’t involve. For example, one of the leading tools in this category had to remove the facial analysis component from its screening assessments, following a 2019 complaint to the US Federal Trade Commission from the Electronic Privacy Information Center and an independent audit. If a process includes automated scoring, the consent notice should tell you.

Conversational screening bots. There are tools that conduct automated screening, interview scheduling, and onboarding, engaging candidates over SMS, WhatsApp, WeChat, and Facebook Messenger. Basically the bot asks knockout questions about availability and work authorisation, then schedules an interview.

Sourcing and matching tools. These help recruiters find people, which affects whether you’re discovered rather than whether you’re rejected. For example, there is LinkedIn’s Hiring Assistant (which is actually not a regular part of every paid LinkedIn Recruiter plan). Beyond LinkedIn, there are also tools that use semantic matching across public profiles and, for technical roles, sources such as GitHub or Stack Overflow. The practical implication for you is the opposite of the myth: these tools mean a well-described profile is more likely to be found, not more likely to be filtered out.

Independent reviewers make the same point: conversational screening works for roles where volume is the main constraint and a CV surfaces little signal, and falls short for specialised or senior positions that need judgment about communication and craft. Games industry hiring sits firmly in that second category. The candidate pools are small, the roles are specific, and pattern-matching doesn’t earn its place.

The Regulation Is Moving Toward Disclosure

Worth knowing, because it shapes what employers can actually do, and because it undercuts the premise of a secret algorithm operating unchecked.

In the EU, the AI Act names recruitment AI directly. Annex III, point 4(a) covers AI systems intended to be used for the recruitment or selection of people, specifically including systems that place targeted job advertisements, analyse and filter applications, and evaluate candidates. The European Commission’s own AI Act Service Desk gives worked examples, including that a system designed to assist recruiters by analysing and filtering candidates should be classified as high-risk.

On timing, the AI Act’s transparency obligations took effect on 2 August 2026. The deeper high-risk obligations for Annex III systems, covering conformity assessment, documentation, and oversight, have been deferred under the Digital Omnibus. Warden AI notes a revised date of 2 December 2027, approved by the European Parliament but, at the time of their writing, awaiting formal Council adoption and publication in the Official Journal. If this affects you commercially, verify the current position with counsel rather than relying on any blog (including this one – we’re not in the position to give legal guidelines).

One point in the Commission’s guidance is worth highlighting for candidates: a tool can be high-risk where it materially influences an employment decision even if a human makes the final call. The presence of a human reviewer doesn’t automatically remove the obligations.

The direction of travel is consistent: disclosure, consent, bias testing, and documented human oversight. Which means that if a serious employer is using AI to evaluate you, they increasingly have a legal obligation to tell you so.

What This Means for Your CV

Here’s the practical part, and it’s less exciting than the mythology suggests.

Formatting matters, but for a boring reason. Parsing is the one place where software genuinely trips over CVs. Text inside images, complex tables, unusual multi-column layouts, and headers or footers containing essential information can all confuse the extraction step. Not because a robot dislikes your design, but because it can’t read it. A clean single-column layout with standard section headings and a PDF export solves this in about five minutes. Then stop thinking about it.

Use the job description’s vocabulary, honestly. If a posting says “Unreal Engine 5” and your CV says “UE5,” a keyword search may not connect them. Write out both where it’s natural. This isn’t gaming an algorithm, just making sure a recruiter’s search finds you. The line is honesty: describe what you’ve actually done, in the words the industry uses for it.

Don’t keyword-stuff. White text on white backgrounds, hidden keyword blocks, and pasting the entire job description into your CV in tiny font are all things people genuinely try. They don’t work, they look absurd when a human opens the file, and they’re a fast route to being remembered for the wrong reason.

Answer screening questions carefully. This is the actual automated gate. Rushing a visa or location question is far more likely to end your application than any formatting choice.

Then write for the person. The recruiter reading your CV is looking for evidence you can do the job: what you shipped, what you owned, what tools you used, what changed because you were there. That’s what gets you an interview. It was always what got you an interview.

---

Common Questions About ATS and AI in Hiring

Where does the “75% of CVs are rejected by ATS” statistic come from?

A 2012 sales pitch by Preptel, a company selling resume-optimisation services. No methodology was ever published, and the company ceased trading in 2013. Career consultant Christine Assaf traced the claim and found no study or survey behind it, as documented by HiringThing and The Interview Guys. It has been repeated widely enough since that it now gets cited as established fact, but there is no research supporting it.

What are knockout questions and can they reject me automatically?

Yes, and they’re the most common form of genuine automated filtering. They’re yes/no questions attached to an application form, with a recruiter specifying which answer disqualifies. Typical examples are visa status, current location, and willingness to work on-site. It’s a rule someone wrote, not an AI judgment. Answer them carefully, because a rushed answer here will close your application faster than any formatting issue.

Does 8Bit use AI to screen applications?

No. We don’t use an AI-based applicant tracking system, and every application that reaches us is read by a recruiter. The only automation involved is knockout questions on platforms that impose them, such as LinkedIn’s screening questions covering visa status, location, and on-site availability. We don’t use screening questions on our own website.

Should I optimise my CV for an ATS?

Optimise for parsing, not for a score. Use a clean layout, standard section headings, and a PDF export, so the system can extract your information correctly. Use the vocabulary the job description uses where it honestly describes your experience, so recruiter searches find you. Beyond that, write for the human who will read it. Chasing a compatibility score from a CV-scanning tool isn’t the same as being a strong candidate.

Do any hiring tools genuinely use AI?

Yes. Some ATS platforms offer AI fit scores that rank applicants. Video interview platforms can score recorded responses, working from the transcript of what you say rather than facial analysis. Conversational screening bots handle qualification and scheduling at high-volume employers. These cluster around large-scale and entry-level hiring rather than specialist recruitment.

Why did I never hear back about my application?

Most often, volume and timing. A well-known game studio posting an attractive role can receive hundreds of applications within days, and a recruiter working through that manually is the real bottleneck. Applying early, being genuinely well-matched to the requirements, and being findable through a specialist recruiter all help more than reformatting your CV.