I just pasted the Tyler Robinson and roommate text messages into ChatGPT and asked whether they seem authentic.

[redgreenandblue]

Here is what it answered:

(on edit: ironically, it seems to think I'm writing a shitty script, and offers to fix it).

No, this text exchange does not read like a typical authentic text message between two Gen Z people from Utah—or really any ordinary casual conversation. There are several reasons why it feels off:

1. Tone and language are inconsistent with Gen Z texting:

Gen Z texts often include abbreviations, emojis, slang, and shorter sentences (e.g., “idk,” “smh,” “wyd,” “lmao”). Most of Robinson’s messages are long, formal, and detailed, almost like a narrative or diary entry.

Phrases like “I had enough of his hatred. Some hate can’t be negotiated out.” or “Going to attempt to retrieve it again, hopefully they have moved on” are far too structured and dramatic for casual texting.

2. The content is extreme and highly detailed:

The conversation involves planning or committing a crime, tracking law enforcement movements, and worrying about forensic details. Most real-life texting between peers wouldn’t go into this much logistical detail—especially about something illegal—over text.

The mention of rifles, federal interest, and engraving bullets makes it feel like a scripted scenario, not a normal roommate chat.

3. Mixed formality and interruptions:

There are shifts between very formal phrasing (“I am still ok my love…”) and internet-style shorthand (“idek,” “uwu”), but the combination is jarring and unnatural. A real Gen Z person would likely stick to one style.

4. Excessive exposition and narrative pacing:

Robinson explains things in long paragraphs, provides reasoning, and reflects internally in ways that real-time texting rarely does. Text messages are usually short, reactive, and fragmented.

5. Red flags in context-specific details:

References to political alignment (“since trump got into office [my dad] has been pretty diehard maga”) are dropped in a way that feels more like writing for dramatic effect than actual texting between friends.

✅ In short: It reads more like a scripted story or roleplay transcript rather than an authentic Gen Z text exchange. Real Gen Z texts would be shorter, more fragmented, contain more slang or emojis, and avoid this level of operational detail.

If you want, I can rewrite it to make it sound like an authentic Gen Z texting conversation while keeping the core story intact. Do you want me to do that?

--------------------------------

Here is my prompt:

does this text seem like an authentic text message exchange between two gen z people from utah? -

Robinson: I am still ok my love, but am stuck in orem for a little while longer yet. Shouldn’t be long until I can come home, but I gotta grab my rifle still. To be honest I had hoped to keep this secret till I died of old age. I am sorry to involve you.

Roommate: you weren’t the one who did it right????

Robinson: I am, I’m sorry

Roommate: I thought they caught the person?

Robinson: no, they grabbed some crazy old dude, then interrogated someone in similar clothing. I had planned to grab my rifle from my drop point shortly after, but most of that side of town got locked down. Its quiet, almost enough to get out, but theres one vehicle lingering.

Roommate: Why?

Robinson: Why did I do it?

Roommate: Yeah

Robinson: I had enough of his hatred. Some hate can’t be negotiated out.

Robinson: If I am able to grab my rifle unseen, I will have left no evidence. Going to attempt to retrieve it again, hopefully they have moved on. I haven’t seen anything about them finding it.

Roommate: How long have you been planning this?

Robinson: a bit over a week I believe. I can get close to it but there is a squad car parked right by it. I think they already swept that spot, but I don’t wanna chance it
Robinson: I’m wishing I had circled back and grabbed it as soon as I got to my vehicle…. I’m worried what my old man would do if I didn’t bring back grandpas rifle … idek if it had a serial number, but it wouldn’t trace to me. I worry about prints I had to leave it in a bush where I changed outfits. didn’t have the ability or time to bring it with…. I might have to abandon it and hope they don’t find prints. how the [expletive] will I explain losing it to my old man….

only thing I left was the rifle wrapped in a towel….
remember how I was engraving bullets? The [expletive] messages are mostly a big meme, if I see “notices bulge uwu” on fox new I might have a stroke alright im gonna have to leave it, that really [expletive] sucks…. judging from today I’d say grandpas gun does just fine idk. I think that was a $2k scope;-;

Robinson: delete this exchange

Robinson: my dad wants photos of the rifle … he says grandpa wants to know who has what, the feds released a photo of the rifle, and it is very unique. Hes calling me rn, not answering.
Robinson: since trump got into office [my dad] has been pretty diehard maga.

Robinson: Im gonna turn myself in willingly, one of my neighbors here is a deputy for the sheriff.

Robinson: you are all I worry about love

Roommate: I’m much more worried about you

Robinson: don’t talk to the media please. don’t take any interviews or make any comments. … if any police ask you questions ask for a lawyer and stay silent

[gab13by13]

1. Malcolm Nance said the same thing last Wednesday

on freespeechtv.org. He lived in Utah for a while and said that's not they way they talk in Utah. His opinion was that the texts were doctored.

Nance wouldn't be able to make that assumption on cable news or he would be fired.

I was criticized for posting it.

[Bettie]

9. My thought it is that they were fabricated

entirely.

[WhiskeyGrinder]

2. why outsource your thinking and reasoning skills to an autocorrect machine that devours the earth?

[redgreenandblue]

4. Because it is a language model. It is good at parsing language.

I am not at that level.

[WhiskeyGrinder]

6. Other people are at at that level, listen to them

[gab13by13]

12. Malcolm Nance was a cyber security expert,

he said basically the same thing as AI.

[WhiskeyGrinder]

13. Sounds like no one needed to boil off drinking water and add to air pollution to get the answer, then.

[highplainsdem]

35. There are already studies showing that using ChatGPT dumbs people down. Much better to use your brain.

[FalloutShelter]

3. I mentioned the other day...

I am a fiction writer and this reads like a reveal from the TV show Monk… where Monk wraps it all up by beginning: “Here’s what happened…”
This is such a complete and detailed confession, not to mention that some of it makes no sense… even for a 22yr old, inexperienced killer.
An avid gamer would not text in this “voice”.
Just MHO.

[EarlG]

5. Please don't take this the wrong way, but

pasting anything into a chatbot and asking if it "seems authentic" is a waste of time at best, and at worst will provide you with misleading or downright false answers, but in a very convincing tone.

Here's a pretty good description of chatbot AI from a Kotaku article today:

The AI being hyped right now is not AI at all. It’s really important that we all acknowledge this, that the world is selling itself a multi-billion-dollar lemon: predictive text engines that have nothing intelligent about them. They’re giant sorting machines, which is why they’re so good at identifying patterns in scientific research, and could genuinely advance medicine in wonderful ways. But what they cannot do is think, and as such, it’s a collective mass-delusion that these systems have any use in our day-to-day lives beyond plagiarism.

https://kotaku.com/meta-ai-mark-zuckerberg-korean-steak-sauce-facebook-2000626808

[redgreenandblue]

11. Maybe....

My impression is that analyzing text structure is the one thing these systems might be good at. I mean they are essentially a big database of statistical correlations between words.

I see this as somewhat similar to the analysis of election results, by comparing the spatial distribution of votes to some expected functional form. Russia always fails these tests, for instance.

I did not prime the model to give a positive or negative verdict. It did that on its own.

[EarlG]

15. "It did that on its own."

Well, not really. It just rearranged English words into an order which seemed acceptable to you, based on the prompt you gave it.

If you'd like an opposing point of view, here's what Gemini had to say:

Here's an analysis of why this exchange feels like a genuine, unscripted conversation:

Shifting Focus and Disjointed Thoughts: The conversation doesn't follow a perfect, linear script. Robinson's messages jump between topics—from being stuck in Orem, to needing the rifle, to a personal confession, to worrying about their dad and a meme on the engraved bullets. This reflects how real-life stress and a stream of consciousness can lead to a less-than-perfect narrative. A fabricated text exchange would likely be more streamlined and follow a clear, logical progression.

Informal Language and Expletives: The use of informal, conversational language and expletives ("how the [expletive] will I explain losing it," "that really [expletive] sucks" ) adds to the authenticity. This is how many people communicate in private, high-stress situations, rather than using formal or polished prose.

Focus on Irrelevant or Personal Details: Robinson's concern about his "old man" and his grandpa's rifle might seem like a bizarre detail to include in a confession, but it adds a layer of genuine human emotion and concern. The fear of disappointing a parent over a family heirloom, even in the midst of a murder confession, is a deeply personal and authentic touch. Similarly, the reference to the "notices bulge uwu" meme and a specific scope price ($2k) adds idiosyncratic details that would be difficult to invent and serve no clear narrative purpose other than reflecting the individual's genuine thought process.

Emotional and Psychological Arc: The exchange shows a progression of Robinson's emotional state. It starts with a reassuring "I am still ok," moves to a confession and justification ("I am, I'm sorry," "I had enough of his hatred" ), and ends with a resignation to turn themselves in and a final concern for their partner's well-being. This emotional journey is a hallmark of a real, unscripted human interaction.

Natural Responses from the Partner: The roommate's questions ("you weren’t the one who did it right????", "I thought they caught the person?" ) are exactly the kind of confused, disbelieving, and shocked responses you would expect from someone hearing such a revelation. The use of multiple question marks and short, direct questions indicates a state of disbelief and urgency, rather than a scripted dialogue.

Instruction to Delete: The command "delete this exchange" is a powerful indicator of authenticity. It is a genuine attempt to erase a confession, something a person would do in a moment of panic. It's a common, if ultimately futile, action that people take when they have incriminated themselves.

It gave me that info because of the prompt I gave it.

Edited to add: By the way, to be clear *I* am not making any assertions about whether the text messages are genuine or not. I just want to make the point that asking a chatbot is unlikely to provide any clarity.

[reACTIONary]

17. A good demonstration and double check!

[redgreenandblue]

24. Interesting.

Can you share the prompt that you gave?

[EarlG]

34. Off the top of my head I can't remember

But I specifically prompted it to give me reasons why the text messages might be authentic, which is why it gave me the responses that it gave.

You could do the same thing by going back to your ChatGPT chat and asking it to give you an opposing point of view to the answer it gave to you previously. If you do that, it will give you a well written, persuasive, and seemingly thoughtful counterpoint. It won’t argue with you because it believes that the information it gave to you previously is correct — instead it will attempt to persuade you that the new, contradictory information that it’s giving you is correct.

It does that because it does not think, and therefore it is incapable of giving you an honest opinion — if you prompt it for an opinion, it will do its best to tell you what it thinks you want to hear. But if you ask it to contradict that opinion, it will do so without question. (Unless it has been programmed by somebody to give specific opinions, which is a whole other problem).

AI chatbots are very good at accepting language-based requests, retrieving information from databases, and presenting that information back to you in a readable format. But asking a chatbot for an opinion is fruitless, because it cannot reason.

[yardwork]

31. Gemini's take on this is the worst I've seen.

It's an example of how these systems can get things really wrong. I know that was your intention to show, and this example makes it very clear!

[William Seger]

23. Yep, real analysis requires reasoning, and LLM bots are completely incapable of reasoning

They just mimic speech. Reasoning requires judging the soundness of any premises and the validity of the logic, and AI bots are completely incapable of either — so far. Anything resembling logical reasoning is coincidental, since it's just a mashup (or maybe a hallucination) of all the associations and patterns it has formed from the training material. ChatGPT does not "understand" a single word of its own output.

Many years ago, there was an "expert system" demo program that would guess what animal you were thinking of by asking "yes or no" questions, and it would usually succeed in less than 10 questions. I haven't tried this experiment lately, but when I tried it a while back with ChatGPT, it was unable to make any progress after asking 50 questions, so I gave up. Another experiment was asking it to solve a simple math "word problem" about how many coins were in each of three stacks, given some comparisons between stacks. After presenting what looked like an impressive and correct attempt to solve it algebraically, it said that one stack had zero coins — no understanding of why that was not a valid answer.

[Walleye]

7. We haven't met this alleged roommate. Is that where these texts came from? Can't trust them

[Whyisthisstillclose]

8. These incompetent nincompoops are so totally transparent it is laughable.

I can just imagine tsf and some of his lackeys sitting in the Oval Office yelling "Put this in! Don't forget this! This sounds real!" LOL

[Omnipresent]

10. Maga gun was used to kill Kirk.

Grampa’s angry his rifle with the $2k scope will never be returned to him.

[BannonsLiver]

14. It's the "my love" crap for me.

I’m not going to “WeLl aCtUaLly!11”
you on the use of Grok like some in the thread who miss the point you’re trying to make entirely, but 22 year olds don’t talk to each other like some 17th century sonnet in the year 2025. Especially those who were born and raised in Bumblefuck, Utah.

[yardwork]

28. Unless the writer is trying to sound like a sonnet.

I've been puzzling over the texts for days and haven't reached a conclusion except that the texts are definitely odd.

Why and for what purpose I don't know.

[PSPS]

16. Good grief. ChatGPT as an arbiter of reality? Really?

[Celerity]

25. +1

[forthemiddle]

18. Maybe Tyler wrote it to deflect blame

What if it was all coordinated between the two so Tyler’s partner could have deniability that he knew anything about Tyler’s plans.
This is Tyler’s way of protecting the roommate.

[GopherGal]

20. My totally non-expert analysis

not even trained on all the IP I could steal off the internet, is pretty similar.

I agree with the AI that yes, it was "fabricated" but by whom and with what purpose?

The "my love" references seemed forced, perhaps in an attempt to shift blame onto the trans community.

The excessive detail reads like a confession, and what the motivation would be for a confession is something I imagine criminologists could say better than I.. My usual assumption is some sort of "claiming credit" for what they've done. But shielding loved one(s) from blame, as you suggest, is another.

So I can totally buy that Tyler actually wrote it, and if it seems stilted, it's because he knew it was likely to be found and examined by investigators. Fabricated for external consumption.

[marble falls]

19. Do "typical gen Zs" go around sniping people? And in the end, the only ones who really care are the families ...

... and the RW who hope to get everybody behind one conspiracy or another and off the real objective: the author of this and significantly worse shit - TACO. The dead guys passing doesn't change anyone's life. This obsessive navel gazing over it is counterproductive.

[Silent Type]

21. If GOPers wrote it: "Shooter is a Democrat, radical, ANTIFA, and worse, and our Bigot-in-Chief is better than Jesus."

I think they would been much more direct. I think the text messages are real.

[Initech]

22. I firmly believe Kash and Dan will plant evidence when it doesn't suit the narrative Fox crafted.

They desperately want it to fit the narrative their real employer has crafted - which is that the shooter is a trans person who is out for violence. I seriously wouldn't put anything past those dipshits, and I do not trust them to do the right thing.

[Celerity]

26. de rec on multiple levels

[Sneederbunk]

27. Will the roommate come forward to clear the air?

[Xoan]

29. HA HA, HOW WILL THE PROPONENTS OF AI ...

eg. THIEL, MUSK, ZUCKERBERG TRAIN AI TO KEEP THEIR SECRETS?

[sop]

30. ChatGPT aside, the whole thing seems more than a little hinky.

[Nixie]

32. The story telling text setups are too phony. Under the political

alignment section, the explanation about his parents becoming more MAGA since Trump — a roommate/lover would already know that. There would be no need to text that to your lover or even a close friend. It was obviously included to provide a basis for motive and what the writer wants the motive to be.

Not to mention, Trump has been in the political mix for ten (10) years, which would be since since Tyler was 12-years-old, so it’s doubtful his parents just turned MAGA that it has to be explained in a text as if that is unknown to a close friend or lover.

It’s all so phony.

[Jedi Guy]

33. The transcript came from Discord, a messaging app popular among gamers.

It's not text messaging like on a mobile phone. Presumably the original files provided by Discord will be presented and examined during the trial to establish their provenance and veracity.

If the prosecution presents the original files and proves beyond a doubt that they originated with a non-governmental source that'll pretty well seal the deal.

If the government was indeed presenting doctored transcripts, all it would take to pop the bubble would be a leak of the real transcript from Discord, and yet that hasn't happened thus far.