> ## Documentation Index
> Fetch the complete documentation index at: https://docs.covver.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompting tips

> Practical ways to get better pages, issues and edits from the Covver Assistant.

The Covver Assistant works best when you understand how it builds a page. It does not draw a picture of a layout. It writes the content as blocks (a headline, an intro, body text, a photo, a caption, a pull quote) and picks a page style; a layout engine then measures the text and arranges everything on a grid. Your prompt has the most effect on the content, the kind of page and the look, and less on exact positions.

## Say what kind of page you want

Name the page type. The assistant knows many page styles and picks one from your words.

| Family | Examples you can ask for |
| - | - |
| Covers and closing | a cover, a poster-style cover, a fashion cover, a back cover |
| Front matter | a table of contents, an editor's letter, credits, contributors |
| Articles | a feature with a big photo, a two-column article, an interview, a recipe |
| Photo pages | a full-page photo, a photo grid |
| Lists and data | key numbers, a list, a product grid, a timeline, an events page, a comparison |
| Special | a quote page, a section opener, an ad |

"Make a page about our summer menu" works. "Make a recipe page for grilled peaches with a full-width photo on top" works better.

## Give it the content

The assistant writes copy if you do not provide it, but your own facts make the page yours.

* Paste the real text, names, dates, prices and quotes you want on the page.
* Say how long the text should be: "a short intro and three paragraphs", "headline of five words or fewer".
* Say the language if it differs from the one you are writing in.

> [!TIP]
>
> Keep headlines short. The layout engine sizes type to fit the space, so a long headline gets smaller and a long article may be trimmed. If text was cut, ask for a shorter version or split it across two pages.

## Be clear about add, change or replace

The assistant first decides whether you want to add something, edit what is there, redesign the page, or start the page over. Make that obvious.

| You want | Say something like |
| - | - |
| To add to the page | "Add a pull quote under the second paragraph." |
| To change existing things | "Make the headline bigger and change it to dark green." |
| A new design, same content | "Redesign this page, keep the text and photos." or "Show me another option." |
| A brand-new page | "Replace this page with a full-page photo cover for the autumn issue." |

If a request is ambiguous on a page that already has content, the assistant leans towards adding rather than replacing, so it does not wipe your work.

## Select before you ask

For edits, click the element (or several) first. Requests like "shorter", "bolder" or "translate this" then apply to exactly what you selected, with no need to describe where it is.

## Describe the look once, for the whole magazine

Ask for the style at magazine level and the assistant reuses it on later pages:

* "Use deep blue, warm cream and one orange accent across the whole magazine."
* "Serif headlines and a clean sans-serif for body text, on every page."

When you add pages later, say "in the same style" to keep them consistent.

## Use templates when you want a proven design

Ask for a template by kind: "use a template for the table of contents". The assistant fills a curated template's text and photo slots with your content. Template text slots have a designed length, so content close to that length fits best. You can also browse and apply [templates](https://docs.covver.com/builder/templates.md) yourself.

## Plan whole issues in one request

For a full magazine, give the topic, the audience, the number of pages and anything that must be included:

> An 8-page travel magazine about Lisbon for first-time visitors: cover, contents, a neighborhood guide, a food feature, a two-page photo spread, a day-trip list and a back cover. Warm, friendly tone.

The assistant plans the pages before building. Read the plan and correct it there; changing a plan is faster than fixing eight finished pages.

## Steer the photos

* Describe subject, setting and mood: "a close portrait of a ceramic artist in her studio, soft window light".
* To use your own images, say "use my photos" or "from my library".
* Do not ask for words inside a generated photo. Put text in text elements, where you can edit it.
* To adjust a photo you already have, select it and describe only the change: "same photo, but make the sky stormy". See [AI image editing](https://docs.covver.com/ai/image-editing.md).

## Show it a reference

Attach an image of a layout you like and say what you want from it. With no extra text, the assistant recreates the image as editable layers. With instructions ("use this as inspiration for the cover, but with my title"), it treats the image as a reference.

## Iterate in small steps

* Fix one thing per message: first the headline, then the photo, then the colors.
* If a result is close, edit it by hand. Everything the assistant places is a normal element.
* If a result is far off, undo and rephrase rather than stacking corrections.
* Start a **New chat** when you switch to an unrelated task, so old instructions do not carry over.

## What prompts cannot do

* Exact positions and sizes ("12 pixels from the left") are faster to set by hand.
* The assistant does not publish, and it does not generate video or audio.
* It cannot look up live facts for you. Provide the real names, numbers and dates you want printed.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.