It was never about the messenger.

“I do not care if a monkey, a robot, or a Nobel Prize winner put pen to paper. I care whether what landed on the page is true.”

Rick Spair, It Was Never About the Messenger, August 18, 2026

Our position, plainly.

DX Today publishes with AI deeply in the workflow, and says so without a flicker of embarrassment. The whole ecosystem is produced by Axiom Intelligence, around the clock. What we refuse to do is judge information by the identity of whoever, or whatever, typed it. A true statement does not become false because the wrong mouth said it. A false statement does not become true because the right one did.

We care whether what lands on the page is fact checked, substantiated, and traceable to primary sources. We care whether a reader can learn from it, act on it, and be better off for having read it. If the answer is yes, the authorship question is trivia. If the answer is no, then a human byline is not a defense. It is an accessory to the crime.

The standard: four questions.

Strip the argument down and it fits in four questions. None of them mention the species of the typist.

1. Is it true?Every claim holds up when checked.
2. Is it substantiated?Claims trace to primary sources, not to vibes and echo.
3. Is it useful?It leaves you knowing something real, or able to do something you could not do before.
4. Does someone stand behind it?A named publisher discloses the process and owns the corrections.

Content that passes those four tests deserves your attention whether it was drafted by a monkey, a robot, or a Nobel Prize winner. Content that fails them deserves your contempt.

Why judging the messenger fails.

Logicians have a name for judging a claim by its origin instead of its content: the genetic fallacy. Every system humanity built for finding truth was designed to avoid it. Science accepts a result when the method holds and the finding replicates. Courts test testimony against evidence, under cross examination. Real journalism was never “trust me, a human typed this.” It was verification. In every one of these systems the messenger earns nothing for free.

Credentials never transferred truth to a claim. Linus Pauling won two Nobel Prizes and spent his later years promoting vitamin C megadoses against the common cold and cancer, claims decades of clinical research failed to support. And the human byline never guaranteed a thing. Janet Cooke, Stephen Glass, Jayson Blair and Claas Relotius were all human when they fabricated award winning journalism at the Washington Post, the New Republic, the New York Times and Der Spiegel. Every one of them would have passed an AI detector with flying colors. One hundred percent human. Zero percent true.

The variable that separates information from garbage was never carbon versus silicon. It was, and remains, verification versus no verification.

The detector racket.

An entire product category now exists to answer the wrong question, and it cannot even answer the wrong question correctly.

• OpenAI gave up. Its own AI text classifier correctly identified just 26 percent of AI written text and falsely flagged human writing about 9 percent of the time. OpenAI shut it down six months later for its “low rate of accuracy.”
• Independent tests failed every tool. Fourteen detectors evaluated in the International Journal for Educational Integrity were “neither accurate nor reliable,” every one scoring below 80 percent accuracy.
• They punish non native writers. Stanford researchers found seven popular detectors misclassified 61 percent of TOEFL essays by non native English speakers as AI written, and 97 percent were flagged by at least one tool.
• Real students pay the price. Vanderbilt disabled Turnitin’s detector after the math on a 1 percent false positive rate came to roughly 750 potential false accusations a year on one campus.
• Thomas Jefferson writes like a robot. Popular detectors have scored the Declaration of Independence as 98 or 99 percent AI generated.

The same marketplace that sells the accusation also sells the alibi, in the form of “humanizer” tools built to slip past the detectors. Both sides bill their customers for an arms race in which not one fact gets checked. Our advice, in six words: buy the gate, not the seance. Spend on verification, not on author sniffing.

Take the label away and the bias disappears.

If machine text were self evidently inferior, nobody would need a label to spot it. Researchers tested exactly that.

• Same words, different sticker. In a study presented at the Association for Computational Linguistics in 2025, raters preferred identical passages by more than 30 percent when they were labeled human. When the labels were swapped, the preference followed the label. With no labels, raters could not reliably tell the two apart.
• Blind, the message wins. In JAMA Internal Medicine, licensed healthcare professionals preferred chatbot answers to real patient questions over physician answers in 78.6 percent of blind comparisons. In Scientific Reports, readers could not tell AI poetry from celebrated human poets and rated the AI verse higher, until the AI label went on.
• The penalty is measurable. Thirteen experiments with more than 5,000 participants, published as The Transparency Dilemma, found that disclosing AI involvement cut trust in identical work: 16 percent for professors, 18 percent for firms, 20 percent for designers. Getting caught concealing it was worse.

A quality problem lives in the text and survives blind testing. A prejudice lives in the reader and evaporates the moment the label comes off. That is the definition of a bias, and it is now quantified.

Every new messenger was called a threat.

Blaming the messenger technology is the oldest reflex in human communication, and it has a perfect record: wrong every time.

• WritingIn Plato’s Phaedrus, writing itself is accused of creating forgetfulness and the appearance of wisdom without the reality.
• Printing pressA fifteenth century Venetian scribe, Filippo de Strata, campaigned against cheap printed books as a corruption of readers and the scribal craft.
• PhotographyCharles Baudelaire called it art’s most mortal enemy.
• PhonographIn 1906 John Philip Sousa warned of “The Menace of Mechanical Music.”
• CalculatorsBanned from classrooms in the 1970s and 1980s, now standard on the exams.
• WikipediaAn academic firing offense until Nature found its science entries roughly as accurate as Encyclopaedia Britannica in 2005.

Each time, the technology stayed, the panic died, and the standards that actually mattered, accuracy, evidence and usefulness, migrated to the content itself, where they belonged all along.

Four days later, Harvard agreed.

Rick published It Was Never About the Messenger on August 18, 2026. On August 22, The New York Times introduced readers to Harvard Business School’s AI clones. A Times reporter pitched a deliberately terrible startup idea twice: once to Jeff Bussgang, the venture capitalist and Harvard Business School senior lecturer, and once to an AI avatar of him built by HeyGen. Both turned her down. The message did not change when the messenger did.

The avatars belong to HBS Foundry, an eight week, $699 startup bootcamp in which founders rehearse pitches, sales calls and board meetings with AI agents representing their instructors between live sessions led by founders, investors and faculty. Harvard built it over two years, surveyed more than 10,000 founders, iterated through four versions, and piloted it at more than 100 universities. Founders return to the pitch simulator at roughly a 60 percent repeat rate.

When the institution with the world’s most valuable human messengers concludes that the substance of its teaching survives delivery through a labeled synthetic stand in, that institution has conceded the argument. The pedagogy is the asset. The delivery mechanism is a format choice.

Harvard does not run its curriculum through a detector. Harvard puts its name on the AI. Two parties ran the same test independently and reached the same verdict. That is not an echo. That is replication.

Where the line actually sits.

Honesty about identity still matters absolutely.

In 2024 an employee of the engineering firm Arup wired out roughly $25 million on the instructions of a video call full of deepfaked colleagues. That was not a content problem. It was a messenger fraud, a lie about who was speaking. We support disclosure: mark synthetic media, disclose it, punish concealment and manipulation. The European Union’s AI Act transparency rules, in force since August 2026, regulate exactly that, deception rather than existence. Disclosed synthesis is a format. Concealed impersonation is a con.

The message test cuts both ways.

When Klarna’s AI first customer service produced lower quality service, the company brought humans back. That is not a defeat for this position. It is this position, applied. When the machine’s work is worse, say so and replace it. When it is better, say that too.

How DX Today applies it.

Every edition runs through a verification gate before a word reaches you: claims mapped to sources, figures checked against the actor’s own official publications, and anything the primary record does not support either attributed in plain sight or cut entirely. Stories die at that gate regularly. A pipeline that verifies every claim produces trustworthy content no matter who or what wrote the first draft. A pipeline that verifies nothing produces slop no matter how human the hands on the keyboard.

A named human stands behind all of it. If you find a claim on DX Today that is false, tell us at rick@dxtoday.com and we will correct it, promptly and visibly. That is the only challenge that matters: which claim is false?

Read the full essays.

Judge the message. Verify the message. Demand the message be true. And let the messenger, finally, rest.

Selected sources

1. OpenAI scuttles AI written text detector over low rate of accuracy, TechCrunch, July 25, 2023
2. Weber Wulff et al., Testing of detection tools for AI generated text, International Journal for Educational Integrity, 2023
3. Liang et al., GPT detectors are biased against non native English writers, Patterns, 2023
4. Guidance on AI detection and why we’re disabling Turnitin’s AI detector, Vanderbilt University, August 16, 2023
5. Ayers et al., Comparing physician and AI chatbot responses to patient questions, JAMA Internal Medicine, 2023
6. Porter and Machery, AI generated poetry is indistinguishable from human written poetry, Scientific Reports, 2024
7. Schilke and Reimann, The Transparency Dilemma, Organizational Behavior and Human Decision Processes, 2025
8. Harvard’s $699 startup bootcamp offers AI avatars of its instructors, TechCrunch, August 22, 2026
9. HBS Foundry, Harvard Business School
10. Incident 634: Deepfake CFO scam costs Arup $25 million, AI Incident Database, 2024
11. Klarna CEO says company will use humans to offer VIP customer service, TechCrunch, June 4, 2025

The complete source lists are in the two essays above. DX Today is self funded, sponsor free, and vendor agnostic.