Which tool to choose
Cutting off halfway happens on all assistants, because each one has a ceiling on how much text it produces in a single answer. The difference is how they react to the command to continue.
- Almost all AIs understand "continue" and pick up the thread on their own. It's always the first thing to try.
- When a plain "continue" makes them restart badly, you need the textual anchor: you indicate the exact point to resume from. It works everywhere and is more reliable.
Don't change tools for this: the cutoff isn't a flaw of a particular AI, it's the physical limit of how much it can write in one go. The remedy is in the way you ask for the continuation, not in the assistant.
How to do it
From a computer or a phone the commands are the same.
- As soon as the answer stops halfway through a sentence, simply write:
continue
- If it picks up well, let it finish; it may take more than one "continue" for very long answers.
- If it restarts from the top, repeats chunks, or seems to have lost the point, use the precise anchor. Copy the last line it had written and paste it. The operational syntax:
Resume exactly from where you stopped, without repeating
and without restarting. Your last words were:
"[paste here the last sentence it had written]"
Go on from there.
- If the answer breaks every time because it's enormous, change strategy: ask for it in pieces from the start. "Write me only the first part, then I'll stop and ask you for the second."
- Feedback: at the end, reread the junction point between the cut-off piece and the continuation. If the sentence stitches together without gaps or repetitions, you've completed it well. If there's a break, re-paste the two parts and ask for them to be joined cleanly.
A concrete example
Giorgio asks the AI for a long article for his site. The answer stops halfway through a paragraph, in the middle of a word. He writes "continue": the AI resumes, but restarts the paragraph from the top, and now there are two versions of the same piece.
Giorgio changes approach: he copies the last words left on screen ("...and for this reason it's worth") and writes "resume exactly from here, without repeating: and for this reason it's worth." This time the AI continues cleanly from that point. Then he joins the two parts by hand. The textual anchor solved what "continue" alone had confused.
When it does NOT work (and how to fix it)
If after "continue" it restarts from the top or repeats
The generic "continue" made it lose the point. Switch to the textual anchor: copy the last sentence written and tell it "resume exactly from these words, without repeating them." Indicating the precise point eliminates the ambiguity that makes it start over.
If it cuts off on every answer, even a short one
Then it's not the length limit: it's a technical hitch. Reload the page, check the connection, try an "incognito" window without extensions. When the cutoff hits even short answers, the suspicion shifts from the content to the connection.
If the parts keep not matching when you join them
When the junction point stays confused, stop chasing the continuation and ask for a targeted redo: "this piece came out broken, rewrite me from scratch only the paragraph about [topic], complete and continuous." Sometimes redoing a whole piece is quicker than stitching together two halves that don't fit.
If you need a very long text and it always breaks
Don't fight the limit: go along with it. Ask for the text in sections from the start, one per message. "Let's do the introduction, then I'll stop." Once you have it, "now the first chapter." Building a long text in blocks is more solid than asking for it all at once and gathering up the pieces.
A tip from someone who actually uses it
For long texts, anticipate the cutoff instead of suffering it: tell the AI from the very first request to proceed in sections and to stop after each one, waiting for your go-ahead. That way it never breaks mid-sentence, because it stops where you decide, at a clean point. Steering the rhythm of the answer is more effective than chasing an answer that has already broken at the wrong spot.
Frequently asked questions
Why does the AI stop right in the middle of a sentence?
Because it has a ceiling on how much text it produces in a single answer, and when it reaches it, it stops wherever it happens to be, even in the middle of a word. It's not an error or a change of mind: it's the wall of the maximum length of an answer. That's why the remedy is simply to make it continue.
Does "continue" work in all languages?
Yes, and it also works writing "go on," "proceed," "complete." The AI understands the intent. If one phrasing doesn't make it resume well, try the textual anchor with the last sentence: it's the most robust method, independent of the exact words you use to ask.
Can I avoid the interruptions entirely?
Not eliminate them, but prevent them: by asking for texts in sections instead of in one enormous block. If every request stays under the maximum length of an answer, the AI has no reason to break off. The cutoff is almost always the symptom of a request too big to fit in a single answer.
If I always have to say "continue," is the AI poor quality?
No, and it's an expectation to correct. The per-answer length limit exists in all assistants, even the best ones, and it's a technical choice, not a quality flaw: it serves to keep answers manageable and fast. An AI that writes a very long text all in one breath wouldn't be smarter, it would just be more awkward to read. "Continue" doesn't measure how good the AI is: it measures how long what you asked for is.