A Recruiter Failed the Bread Test

Names and companies have been anonymized. Quotes are reproduced as-is. The image is not published in its original form: it contains the candidate’s name and the company’s logo.

“I can assure you that you’re communicating with a real person, not a bot.”

That’s what a recruiter wrote to me. Right before giving me his bread recipe.

I found that reassuring. Then he added a warning about his company’s data-retention policy. For a bread recipe. That’s when I started asking questions.

To understand how you end up asking a recruiter for a bread recipe, go back four days.

The message that expires

On November 11, at 18:26, I get an email in Gmail confidential mode. The subject looks like a template: my handle, followed by “Your Experience.” The link expires. I find that a bit theatrical for a first contact, but why not. A recruiter who takes precautions is a good sign.

The message comes from a “Senior Technical Recruiter” at an American firm. He found my profile and has a role for me: Information Security Engineer, Application Security, at a European software company. 100% remote, based in France, permanent contract, €95,000 to €125,000. The role is exactly what I’m looking for. It’s almost embarrassing.

The email mentions SEK, CPG, Joern, Semgrep. I didn’t know recruiters read technical blogs. Maybe it’s the new sourcing. Or a scraper. I’ll keep that in mind.

I send him my CV. He tells me the role is perfect, my profile is perfect, everything is perfect. I should have found that suspicious. At the time, I just found it friendly. Then he asks me to stay online for 25 to 40 minutes while the company reviews my file. I had nothing else to do, so I said yes.

I didn’t know an ATS needed company. Apparently this one wanted me to stick around.

The ATS score

Forty-three minutes later, the verdict comes back: my CV didn’t pass the ATS screening. The recruiter attaches an image. A dashboard, with the company’s logo, my name, the role, and a score: 32%. Apparently a CV can be graded like a used car.

The ATS score on the fake dashboard

This is fine

The recruiter says my CV doesn’t do me justice. That’s nice. He knows someone who can fix it: an “Executive Résumé Optimization Specialist.” It’s paid, of course. The price depends on the level of optimization, and the expert can finish in six hours. The company only accepts two CVs, like a password. I almost asked if it also needed a special character. I send mine again. He tells me the problem isn’t the CV, it’s the optimization, and he brings up his expert again. At this point, I feel like I’m asking for a quote, not interviewing.

The plan

The bread test

I decide to check the recruiter, not the role or the company. I ask him for a bread recipe. I didn’t have a plan. I just wanted to see if he’d answer like a human, or like an assistant who learned to bake between two tickets.

His reply arrives. A recipe. 500 g of flour, 7 g of yeast, 300 ml of warm water, 10 g of salt, a spoonful of sugar, two spoonfuls of oil. Knead for ten minutes. Let it rise for one to two hours. Bake at 200 °C for 30 to 35 minutes.

Then:

“I can assure you that you’re communicating with a real person, not a bot.”

Then:

“Communications conducted through official company recruitment channels may be subject to the company’s applicable monitoring and data-retention policies.”

For a bread recipe. I found the disclaimer a bit disproportionate. The bot said it wasn’t a bot. And then recited its privacy policy. Exactly what a human does when you ask for bread.

The recruiters reply to the bread test

The PNG

I keep the image. I open it. It’s 1536 x 1024, 1.8 MB. The email displayed it at 472 x 315, but the file is bigger. And it contains a chunk most images don’t have: a C2PA manifest. Apparently ATS dashboards sign their screenshots now.

C2PA is a digital ID card attached to an image. It says who generated it, with what tool, and when. It’s signed. It’s usually used to tell a photo from an AI-generated image.

The manifest says:

c2pa.created: 2026-11-12T12:45:37Z
softwareAgent: ChatGPT / gpt-image
digitalSourceType: trainedAlgorithmicMedia
claim_generator: OpenAI Media Service API
signer: OpenAI OpCo, LLC

The image was generated by ChatGPT, not by an ATS. The “ATS score” is just an image. I wasn’t expecting the vehicle inspection to be signed by OpenAI.

The rejection email went out at 14:07 Paris time. The image was created at 12:45:37 UTC, which is 13:45:37 in Paris. Twenty-one minutes earlier. So the recruiter had time to build his dashboard, attach it, and explain that my CV was the problem. That was efficient.

The only ATS in this story is ChatGPT.

Always Has Been

The dashboard is well made, too well made. There’s the company’s logo, my name, the role, the salary, the line “100% Remote (France) - Marseille / PACA Region.” A 32% score, a red gauge, a list of “key issues,” a “Thank You for Applying” button. It looks like a real product.

And the Eiffel Tower in the background. The dashboard says Marseille, the image shows Paris. The model did its best.

The top of the fake dashboard

The “Thank You for Applying” button is the only one that works.

What I checked

The role exists. It was published on November 11, three and a half hours before the first email. The official posting warns that the company only contacts candidates from an address on its domain, never asks for payment, and does not accept CVs from unsolicited agencies. I read it afterward. In hindsight, it’s all pretty clear.

The recruiter uses a Gmail address. He says his firm is submitting my CV. The posting says the company does not accept CVs from unsolicited agencies. I’m not a lawyer, but both versions can’t be true at once.

The recruiter has a LinkedIn profile. It’s old, it has hundreds of connections, a photo, a 2020 banner. So he has a real professional life. But it lists a different firm from the one in his signature. And in the email’s code, the signature image is tagged with a third firm name. Three firms for one recruiter. Maybe a group. Or a one-man band.

I don’t have proof of impersonation, just a pile of clues. A real profile doesn’t prove that the person writing is the person in the profile. Which is a shame. I liked the idea of a recruiter who bakes.

I look around and find LinkedIn posts describing exactly this pattern. A “recruiter” compliments the profile, asks for the CV, announces a bad ATS score, then recommends a paid expert. The comments mention $500. Some say the recruiter and the expert are the same person. The price seemed a bit high for a formatting problem.

The pattern repeats. It’s a product. I got a look at the catalogue, at least.

What to take away

A real recruiter doesn’t ask you to pay.

A universal ATS score doesn’t exist. An ATS reads a CV and compares it to a job description. It doesn’t produce a 32% score in a PNG.

And a human recruiter doesn’t answer a question about bread with a data-retention policy.

I didn’t get the job. I got the recipe. And a quote.

>_ esc