Short answer: because ChatGPT does not look up a stored answer and hand it back. It builds a reply one word at a time, and at every step it picks from a short list of plausible next words rather than always taking the top one. That pick involves a bit of deliberate randomness, so two runs of the same question take slightly different paths. Add saved memory, whatever else is in the chat, and quiet updates to the product itself, and the wonder is that answers repeat as closely as they do. I asked the same question twice and watched what moved To see how far the wobble goes, I gave ChatGPT a small task with rules I could check. In one chat, with the default model selected, I asked for three low-cost ways a small neighborhood bakery could pull in more weekday morning customers, as exactly three numbered items, each under 15 words, no introduction. Then I opened a brand-new chat and pasted the identical prompt. Both replies followed the rules. Neither one used the same sentence...

Short answer: not reliably enough to accuse anyone of anything. I ran two passages through three free AI detectors — one written by ChatGPT, one written by people at NOAA years before ChatGPT existed. All three correctly caught the AI passage. Two of the three also flagged the human passage, one of them at 100%. If a detector has ever been used to judge your work, that is the number that should worry you. The interesting failure with these tools is not that they miss AI text. It is that they accuse real writing, and they do it with a confident percentage attached. How the test was set up A fair test needs two passages that differ in origin and as little else as possible, so I matched them deliberately: The human passage is 179 words from NOAA's JetStream weather education material, about thunderstorm frequency and lightning. It is public domain, written by government staff, and published well before current AI writing tools existed. Provenance is not in question. ...

Short answer: paste both versions into one message, label them clearly, and ask for a numbered list of every difference. On documents that are short and clause-numbered, this works well — I planted six changes in a supply agreement and both tools I tested found all six. What it will not do is tell you which of those six actually matters, and that is the part you cannot outsource. I wanted a test with a right answer rather than a vague impression, so I wrote a fictional eight-clause wholesale supply agreement for an invented bakery and made a second version with exactly six changes in it: four edited values, one clause deleted, one clause added. Then I pasted both into ChatGPT and Claude with the same instruction and counted what came back. What was planted, and what came back Change planted in version B ChatGPT Claude Delivery days: Tue/Thu → Mon/Wed/Fri Found Found Price per case: $42.00 → $44.50 Found Found Payment terms: Net 30 → Net 45 Found Found T...

Short answer: yes. ChatGPT , Claude and Gemini will all accept an image — a screenshot, a phone photo of a printed page, a scan — and read the text in it. The catch is that reading the words and understanding the document are two different jobs, and these tools are far better at the first than the second. What follows is what to send, how to ask, and the specific places the answer goes wrong while sounding completely sure of itself. Where image reading holds up, and where it slips The kind of picture you send matters more than which tool you send it to. Here is what each kind tends to do, and the handling that keeps it out of trouble. What you send How it usually goes What to do about it Screenshot of text on screen The most reliable case by a distance, punctuation and odd spellings included Nothing special. Just ask. Typed page photographed flat Holds up when the whole page is in frame and in focus Fill the frame. One page per photo. A table or a spreadsheet gra...

Short answer: on clean audio — one person, close to a decent microphone, quiet room — the better transcription tools now get roughly 95 to 98 percent of words right. On the recording you actually have, which is three people on a video call and someone dialling in from a car, expect somewhere between 80 and 92 percent. The first number is the one on the vendor home page. The second one decides whether the transcript saves you any time. What 95 percent accurate costs you Transcription accuracy is the flip side of word error rate , usually written WER. It counts three kinds of mistake against the number of words actually spoken: words swapped for the wrong word, words dropped, and words the tool added that nobody said. Ninety-five percent accurate means a WER of about five percent. An hour of ordinary conversation runs to something like 8,000 spoken words. Five percent of that is 400 wrong words, or about one every nine seconds. Spread evenly, that would be survivable. It...

Short answer: two separate things are happening, and only one of them looks like forgetting. Inside one long conversation, the assistant is working from a window of text that has a limit, so instructions you gave near the top quietly stop carrying the weight they did an hour ago. Across conversations, nothing carries over at all unless you deliberately put it somewhere that gets read again. Neither problem is solved by a cleverer prompt. Both are solved by keeping your context outside the chat, where you control it. What follows is the habit that replaces re-explaining yourself every morning: a short brief you keep in a note, a place to park it in ChatGPT , Claude or Gemini , and one move to make when a conversation has clearly drifted. Two kinds of forgetting, and they need different fixes The first kind happens inside a single chat. You set a rule early — keep it under 200 words, never use the word "artisanal", always address the reader as "you" — and for a ...

What I found: I asked ChatGPT to predict twelve interview questions from a job posting and to label each one DIRECT if it came from the posting’s own wording. Nine came back DIRECT. I checked all nine against the posting line by line: six held up, one was a generic opener that would fit any job on earth, and two were built from the background I had pasted rather than the posting. The question list is worth an hour of your time. The labels on it are not worth trusting. If you are in a hurry: paste the posting and your own history into an AI chat, ask for the questions an interviewer would actually ask, and use it as a sparring partner rather than a scriptwriter. The trap is letting it write answers for you — it will invent achievements you never had, and you will not notice until you are saying one out loud. What the list actually looked like The first five of twelve questions, with the source tag and the “what they’re really checking” column. Cha...