Generate Infographics with Gemini in Google Docs: Images, Diagrams, and Batch Visual Editing

Google Docs can now directly call Gemini to generate and edit visual content. Users do not have to go to image tools to create materials, then download, upload and rearrange them. Gemini can read the current document context and generate next to the text:

• infographic;
• Flowcharts or conceptual diagrams;
• with pictures;
• A set of visual content corresponding to multiple chapters;
• Modified existing image.

It also supports adjusting scale, style and content using natural language.

Official update: Generate and edit visuals with Gemini in Google Docs

Quick steps to use

1. Open an editable document in Google Docs on the web.
2. Confirm that the account plan supports this feature and enable Workspace smart features.
3. Enter your visual build requirements from the input field at the bottom of the page or from the Gemini sidebar.
4. Describe the visual type, scope to be summarized, audience, size, and style.
5. Check the generated content for facts, figures, labels, and text.
6. Use subsequent prompts to modify scale, color, layout, or individual elements.
7. After completion, add alternative text and use version history to save the reversible state.

This feature will be rolled out gradually starting on July 28, 2026, and may take up to 15 days to become visible to all eligible accounts. Currently only supports the web side.

Which accounts can be used

Google announced support includes:

Organization administrators need to enable Gemini for Workspace in Drive. End users also need to turn on Workspace smart features. If the program conditions are met but the entrance is not visible, it may be that the functions are still being gradually deployed. Don’t judge that there must be something wrong with your configuration just because a button has appeared on someone else’s account.

Organize your documents before you start

Gemini will refer to context, but cluttered documentation can easily lead to cluttered visual results. It is recommended to do four things first.

Use clear titles for chapters

Use Docs’ heading styles instead of just making the text bold. For example:

• Project goals;
• current issues;
• Implementation steps;
• budget;
• key risks;
• Timetable.

Structured headings help Gemini determine which content belongs in the same section.

Write the numbers completely

Don’t just write “it improves a lot” or “costs less.” If you want to generate a chart, you should provide:

• numbers;
• unit;
• time range;
• data source;
• Compare caliber.

Generative models can format data but should not be required to guess missing values.

Mark content that should not go into the image

Documents may contain internal notes, numbers to be confirmed, and private information. Delete, hide, or explicitly exclude before generating. You can write in the prompt:

Only use the “Public Summary” section and do not cite content from the “Internal Discussions” and “Draft Budget.”

Save a version node

Name the current version by version history before batch building or editing. In this way, you can compare the layout before and after the generation, and can also roll back when the batch modification is not ideal.

Generate your first infographic from the bottom input field

In the Gemini input field at the bottom of the document, don’t just write “Generate infographic.” A complete prompt should contain:

1. which part of the content to use;
2. What visual form is output;
3. To whom;
4. What do you want to emphasize;
5. size or proportion;
6. Text quantity limit;
7. style and color;
8. Which facts are not allowed to be added.

For example:

1
2
3
4
5

根据本文“问题、方案和预期结果”三个章节，
在文档顶部生成一张 16:9 横向信息图。
受众是没有技术背景的项目负责人。
使用三栏结构，每栏最多三个要点，保留文中的数字和单位。
颜色使用深蓝、浅蓝和白色，不添加文档中没有的数据或案例。

After generation, check the content first, and then adjust the aesthetics. Factual errors are a higher priority than ugly colors.

How to write prompts when making a flow chart

Flowcharts require nodes, sequences, and branches. Write the process clearly first, and then ask for visualization. Example:

1
2
3
4

把选中的部署流程转为纵向流程图。
节点依次为：准备账号、检查权限、创建测试文档、生成视觉、人工核对、发布。
在“人工核对”后增加判断分支：通过则发布，不通过则返回修改提示。
不要增加未在列表中的工具或步骤。

If there are multiple parallel paths in the original text, it should be stated which ones are parallel and which ones must be executed in sequence. Models cannot reliably infer real business processes from vague narratives.

Modify existing visual content

Gemini not only generates new images, it can also modify existing visual content based on natural language. Common instructions include:

1

把这张图改成 16:9，保留所有文字和数字。

1

保持结构不变，把配色调整为与文档标题相同的深绿和灰色。

1

删除右下角装饰图形，放大中间的三项结论。

1

将图中的英文标签改为简体中文，产品名和 API 参数保持原文。

By modifying only one or two major targets at a time, it’s easier to tell which instruction is causing the problem. If you are asked to change scale, change language, rearrange structure, add content, and delete elements all at once, the model may miss one of them.

Add pictures to multiple chapters in batches

The official update notes support generating or editing multiple visual content at once. Unified rules should be defined before batch tasks. For example:

1
2
3
4

为“准备、执行、验证、恢复”四个章节各生成一张简洁示意图。
所有图片使用相同的线条风格、蓝灰配色和 4:3 比例。
每张图只使用对应章节信息，不跨章节合并数字。
标题使用章节名，正文标签不超过五个。

After batch generation, check each picture one by one. Don’t just look at the overall style. Common batch errors include:

• A certain figure uses data from another chapter;
• The colors are unified but the icon meanings are inconsistent;
• The same term is translated differently in multiple pictures;
• A certain picture is cropped;
• Label font is too small;
• Numbers have missing decimal points or percent signs.

How to check facts in generated pictures

Create a checklist from content to visuals.

Check the numbers

• Compare item by item with the original text;
• inspection unit;
• Check percent sign;
• Check decimal places;
• Check time range;
• Check if the sum is reasonable.

Check the text

• Product name spelling;
• Name of person and organization;
• Whether Chinese characters are deformed;
• Whether there are unreadable small print;
• Whether to add unsubstantiated conclusions.

Check the structure

• Is the process direction correct?
• Whether the branch conditions are reversed;
• Whether the legend and color are consistent;
• whether the image implies false causation;
• Whether key constraints are omitted.

Generated images are not a substitute for original data and text descriptions.

Common issues with Chinese-language infographics

Chinese visual generation may occur:

• Wrong font;
• Mixing of simplified and traditional;
• Technical identifiers are translated;
• Line spacing is too close;
• Punctuation and line wrapping exception;
• Long sentences crammed into a small area.

The solution is not to keep asking for “clearer” but to reduce the text in the picture. Start by changing long sentences into three- to six-word labels, leaving the explanation in the text or figure captions. API names, filenames, and command parameters should be explicitly requested to remain as is.

What admins need to check

Administrators should confirm:

• Gemini for Workspace in Drive is enabled;
• Use is allowed by the organizational unit where the user is located;
• Workspace smart features strategy;
• Whether the feature is planned to be included;
• Data areas and compliance requirements;
• external sharing strategy;
• Whether the generated images can be used in public materials;
• Document access recovery after employee resignation.

Functions available by default do not mean they are suitable for all departments. Teams handling contract, medical, legal, and unreleased financial data should first establish input and review boundaries.

Gemini input field not visible

Check in the following order:

1. Make sure to use the web version of Google Docs.
2. Check if the account is part of a support plan.
3. Confirm that the document has edit permissions and not read-only or comment permissions.
4. Turn on Workspace smart features.
5. Have your administrator confirm the Gemini settings in Drive.
6. Check to see if the feature is still in its incremental rollout period of up to 15 days.
7. Sign out of the wrong multiple Google accounts and then re-enter.

Don’t just look at the browser avatar to judge between a personal account and a business account. You should confirm the actual identity in the upper right corner of the document.

The generate button is there but the task fails

Possible reasons include:

• The document is too long or the scope of the requirements is unclear;
• Insufficient selected content;
• The request contains unsupported content;
• Temporary network or service abnormality;
• Organizational policy blocks;
• Requesting to generate too many images at once;
• The original image format or size is not suitable for editing.

Start by narrowing it down to a chapter, a picture and a clear goal. If the small tasks are successful, gradually increase the scope.

The image was generated but cannot be used

Common manifestations and treatments:

Don’t skip fact-checking just because the image looks professional.

Export and pre-publish checks

Before exporting a document to PDF, Word, or sharing it publicly, check:

• Whether the image is complete under different page widths;
• Are the fonts and labels in the PDF clear?
• Whether the legend goes with the picture;
• Whether the colors are still distinguishable when printed in black and white;
• Do you need to add sources?
• Whether it contains elements subject to copyright or branding restrictions;
• Whether internal data is leaked;
• Whether to provide alternative text.

When publishing, treat generative AI as a production tool, not a source of truth. Research, news, medical, and financial content especially require human review.

Accessibility

A picture cannot convey meaning solely through color. Suggestions:

• Use text, shapes or textures together in the image;
• Maintain adequate color contrast;
• Avoid fonts that are too small;
• Add alt text to images;
• Keep key figures and conclusions in the text;
• Flowcharts provide corresponding textual steps.

Screen readers cannot reliably understand an infographic with only pixel content. An accessible text version is still necessary.

Docs visual generation function questions

Can it be used on mobile phone?

The official current description only supports the web side. Even if the document can be viewed on the mobile terminal, it does not mean that it has the same entrance for generating and batch editing.

Can you automatically turn data into accurate charts?

It can produce a visual representation, but precise data should still come from a verifiable table. For formal reports, it is recommended to retain the original data and the basis for chart production.

Can I modify multiple pictures at once?

Batch generation or editing requests can be made. Unified rules should be defined first, and the correspondence between numbers, labels, and chapters should be checked one by one.

Will the entire document be read?

It can use document context. If you only want to work on a portion, you should select the content and scope it in the prompt, while removing sensitive information that should not be part of the generation.

Make generated visuals genuinely useful

Gemini generates visual content in Docs, perfect for solving the problem of “the text is complete but lacks clear illustrations”. To get stable results, don’t just pursue generating beautiful pictures once. A more reliable process is:

1. First organize the document structure and data;
2. Limit the generation range;
3. Clarify proportion, audience, and visual norms;
4. Check facts and text item by item;
5. Use version history to control batch modification risks;
6. Supplement accessibility notes and sources before publishing.

The image generated in this way is a reliable expression of the document content, rather than an unverifiable decorative image.