You Can Summon an LLM, But You Can’t Print a Damn Transcript
What happens when genius tech forgets the basics (like a "Print" button that actually works)
This is a story that begins, as so many do, with good intentions—and ends with me in a puddle of frustration and absurdity.
Earlier this week, I was working on something useful for my company’s website: an interactive Google Map showing all our project locations. A good idea, I thought. Practical. Elegant. Maybe even a little . . . snazzy.
Since this involved using Google products, I turned to Gemini (Google's AI) to help me figure out how to implement it. And indeed, as one would expect from a competent AI chatbot, Gemini walked me through using a spreadsheet as a base layer for map pins. It even offered the code needed to convert it to the right kind of file AND to integrate it with our firm's website styles. It was all written out clearly, and I planned to share it with the IT team—because I can barely code more than a simple equation in Excel. I thought, "Hey, this is kind of slick."
But then [cue ominous music]: a fatal mistake. I tried to export the conversation.
I wanted to save it—to keep a record, to share with my team, to refer back to later. A simple enough ask. Surely there's a button, right? Right??
There was not.
So I asked Gemini: “Can’t I just print this conversation to PDF?”
That kicked off what turned out to be SEVEN PAGES of discussion with the AI about how to export the damn chat.
And I mean discussion in the Socratic sense: My input was, "That didn't work" and it kept spinning increasingly complex workarounds. We're talking scrollable divs, full-page screenshots, developer tools, OCR software, and browser extensions with names like FireShot and GoFullPage. At one point, it suggested I take a full-length PNG image of the conversation, run it through Adobe Acrobat Pro's text recognition tool, and then use that output to create a searchable, multi-page PDF.
I had a headache at that point, but it wasn't the end.
After all that, I ended up using Ctrl+A (select all), Ctrl+C (copy it), and Ctrl+V to paste the whole thing into Word. It was a full 19 pages of text with no speaker labels and broken formatting. (BTW, seven of those pages were just me and Gemini going in circles about how to fix the problem it couldn’t fix.)
When I asked Gemini why my choices for capturing this whole conversation were limited to multiple copy-and-paste (one for each input and response on the screen), it acknowledged that this was a common issue for users but there was essentially nothing it could do other than reiterate what it had already told me about options.
My next input was, "If this is a common issue for users, can you send a message to the developers that being able to print a whole conversation / session is essential to users?"
The response was the usual polite blather about how it doesn't have that ability but I could report it through the proper channels, or go write an email like it’s 1998.
(Apparently only an AI can give you "NOPE" spelled out in several paragraphs with bullet points.)
I think that's the point at which, in the middle of that mess, I screamed at my husband when he called me for dinner. Not because I was angry at him, but because the whole experience had pushed me to the point of sensory and emotional overload. And all I wanted to do was save a conversation.
You know, just your average Tuesday.
What are we doing here?
We can summon a large language model that writes code, composes poetry, and spins out essays in the voice of a 19th-century mystic. We can speak to machines in plain language and get useful, often astonishing results. And yet, in the year 2025, we still can’t export the conversation we just had with it.
Not cleanly. Not clearly. Not with a single click. Not without reformatting, guessing who said what, or risking repetitive stress injury from all the copy-pasting.
This is not a minor feature request. This is a basic matter of usability. A transcript is not a novelty anymore; it’s a work product.
The people using these tools aren’t just testing them out for fun. We’re writing proposals, designing workflows, solving problems. And when the AI is helpful, we need to save that work in a format that preserves the structure of the conversation.
So let me spell it out.
If any AI developer out there is listening:
We need a one-click export to PDF or DOCX.
The output needs speaker attribution ("User:" and "AI:" will do).
It should be searchable, paginated, and human-readable.
There should be a print-friendly view that strips out UI clutter.
And we should be able to send feedback from inside the chat window, not by hunting down a form on another site like it's 2006.
For the record, I sent a polite, detailed version of this request to OpenAI. Their response came back in under a minute, presumably written by another AI. It was thoughtful, warm, and confirmed that these features are being discussed.
Good.
I mean, I’m still at the same point where I can summon a generative intelligence to help me build a geographic database. I can ask it about logistics, code structure, or literary theory. I can even ask it to comfort me when I’m sad. But I can’t ask it to export our conversation in a clean, structured, shareable format.
It’s a little thing. But it’s not a small thing.
I want my transcript, dammit. And I don’t think that’s too much to ask.