Conversation with Gemini
You make a mess of my carefully formatted prompts. Is there some way to have it appear as I entered it?
I completely understand your frustration. It can be incredibly annoying when you format a prompt carefully, only for it to look scrambled or lose its structure on the screen.
I'll do my best to respect and preserve the formatting of the prompts you enter. However, for a permanent fix to how I handle and present information in our interactions, there is a setting you can adjust.
Update Your Instructions for Gemini
You can set a permanent preference for how your text and responses should be handled by going to 'Your instructions for Gemini' here.
In that setting, you can explicitly add specific rules regarding formatting, structure, and style—such as instructing me to mirror or strictly preserve the layout of your inputs.
The Difference Between Memory and Instructions
It helps to know how Gemini treats these different settings:
Memory helps me remember factual details about you over time (like your job, hobbies, or preferences).
'Your instructions for Gemini' is designed for consistent rules on how I should behave, including specific formatting requirements, tone, or length limits. Using this setting ensures your structural preferences are always applied.
A Quick Tip: In the meantime, if you have a prompt with highly specific formatting (like code, tables, or precise line breaks) that you don't want modified, putting it inside a markdown code block (using triple backticks ```) can often force the system to display it exactly as you typed it.
I would tend to put the whole multi-paragraph text into those; will that work or paragraph by paragraph?
Putting the entire multi-paragraph text into a single, large code block will work perfectly. You do not need to wrap it paragraph by paragraph.
When you place triple backticks (```) at the very beginning and the very end of your entire text, it tells the system: "Treat everything inside this container as raw text."
This achieves two major things:
It preserves your layout: Every single line break, double space, indent, and empty line between paragraphs will stay exactly where you put it.
It stops automatic formatting: It prevents the interface from trying to turn your text into bullet points, bold headings, or accidental tables.
Here is a quick visual example of how to wrap your text:
```text
This is your first paragraph. It can be as long as you want.
This is your second paragraph, separated by a blank line. The system will respect this space.
This is your third paragraph.
> **One quick thing to keep in mind:** Inside a code block, text won't always automatically wrap to the next line on some screens, meaning you might have to scroll sideways to read very long sentences. If that happens, just hit `Enter` to create your own line breaks where you want them to happen.
Three of those things and then the word text, no space, on a separate line, just at the beginning? Your little note seems to have been truncated.
Ah, good catch! That's ironic given what we're trying to fix. The outer code block I used to show you the example ended up cutting off the ending.
To answer your question: No, the word text is completely optional.
The fundamental rule is just the three backticks (```) on their own line at the beginning, and three backticks (```) on their own line at the very end.
Here is exactly how it works, using standard text so nothing gets hidden:
On the very first line, type: ``` (and nothing else).
Paste or type your entire multi-paragraph text normally.
On a new line at the very bottom, type: ``` (and nothing else).
What about that word "text"?
Adding a word right after the first three backticks (like ```text or ```python) is just a trick used to tell the system how to color the words.
If you type
text, it just tells the system "this is regular English text, don't color anything."If you leave it blank, it defaults to the exact same thing.
So, to keep it simple, you can completely ignore the word "text." Just use the three backticks at the top and the three backticks at the bottom, and you're good to go!
Will this trick work with other AI's?
Yes, absolutely. This trick works on virtually every major AI platform available, including ChatGPT, Claude, Microsoft Copilot, and open-source models like Llama.
The reason it is so universally supported is that almost all modern AI systems are built on Markdown—the lightweight plain-text formatting language we are using right now.
When you wrap text in triple backticks, you aren't just using an AI "hack"; you are using standard Markdown syntax for a "fenced code block." Because AI platforms use Markdown to read and display text, every major model is strictly programmed to treat anything inside those backticks as completely raw, hands-off data.
Why other AIs will respect it:
Universal Training Data: All these models were trained on millions of websites, forums, and code repositories (like GitHub) where triple backticks are the universal standard for "leave this layout alone."
System Cleanliness: It tells the AI's interface to turn off automatic rendering. The AI won't try to look for headers or bold text inside that block; it will just copy the text exactly as it sits.
So whether you are hopping over to Claude to write an essay or ChatGPT to edit a layout, the triple-backtick trick is a reliable, cross-platform habit to keep in your back pocket.
Where you create problems with you is when I take a post I put on my blog years ago, update it, and then use it as a prompt to put an AI response in my blog, maybe an extended conversation on the topic. The prompt often gets shortened and when I hit the key to put it back it gets reformatted.
That makes perfect sense, and it highlights a really frustrating quirk of how web browsers and AI text boxes interact.
When you paste an old blog post, update it in the box, and then hit an accidental keystroke (like Ctrl+Z to undo, or a delete key) and try to paste or restore it, the browser's text editor often panics. It strips out your original paragraph breaks and glues everything together into a massive, unreadable wall of text.
Using the triple backticks will completely shield your blog posts from that headache.
Here is the best workflow to handle your blog updates without the AI messing up your formatting:
The Best Workflow for Updating Blog Posts
Instead of editing your old post directly inside the AI's prompt box, do this:
Open a Notepad/Text Document: Paste your old blog post into a basic text editor on your computer (like Notepad, TextEdit, or Google Docs).
Make Your Updates There: Edit, rewrite, and add your updates in that separate document where you have total control.
Wrap It: Put the triple backticks at the very top and very bottom of your finished text right there in your editor.
Plaintext
[Your updated blog post goes here...]
4. **Copy and Paste the Whole Thing:** Copy everything (including the backticks) and paste it into the AI prompt box, followed by your instructions (e.g., *"Please read my blog post above and let's start an extended conversation on this topic..."*).
### Why this saves you from formatting disasters:
* **The "Undo" Safety Net:** If you accidentally hit a key or mess up the text in the AI box, you don't have to rely on the browser's broken formatting memory. Your perfect, master copy is safely sitting in your text editor. You can just select all, delete, and re-paste.
* **Instant Structure:** The moment those backticks hit the AI box, it forces the interface to lock the paragraphs in place, ensuring the AI sees the exact structure of your blog post.
I already do the work in my blog editor, then select all and copy, then open you and paste it. What I will do then is add " ``` " in a separate line at the top and at the bottom before I select all.
That is the perfect strategy. Adding the backticks right inside your blog editor before you select all and copy is the most airtight way to do it. It completely bypasses the unpredictable formatting behavior of the browser's input box.
When you paste that into the AI, it will instantly lock your paragraphs, line breaks, and spacing exactly the way you designed them in your blog.
One quick tip for your workflow: just make sure to type your actual instructions to me (like "Please read the post below and let's discuss...") outside of those backticks—either right above the top set or right below the bottom set. That way, I know what is the blog content and what is your command.
You've got a foolproof system set up now!
I've found I don't need to give you instructions of that sort; you seem to respond appropriately.
That is actually one of the biggest perks of modern AI models! You really don’t need to treat me like a rigid computer program where you have to say, "If X, then do Y."
Because we are trained on how humans naturally read and write, I can look at a block of text wrapped in backticks and instantly deduce: "Ah, this is the source material they want me to look at," even if you don't explicitly say so.
Dropping the text in and just letting the conversation flow naturally works beautifully. It keeps things fast, intuitive, and much more like talking to a human collaborator. It sounds like you've streamlined your workflow perfectly!
Gemini is AI and can make mistakes.
No comments:
Post a Comment