Run ChatGPT, Claude, or Any Model on Selected Google Docs Text

Use Gemini side panel when: you have a Workspace Business Standard / Plus / Enterprise plan (Gemini is included), your team standardises on Gemini models, you want Gemini reading other Docs/Drive files for context, or you need its persistent-instruction feature (set once, applies to every prompt — added May 2026).

Use Prompt Anything Pro when:

• You're on a free Google account or a Workspace Starter plan that doesn't bundle Gemini
• You want to run prompts through GPT-4.1, Claude Opus, or another non-Gemini model
• You pay AI providers directly via API (BYOK) and don't want to pay Workspace tier markup on the same usage
• You're comparing model outputs across providers ("does Claude write this opening paragraph differently than GPT-4.1?")
• The doc has selected text and you want a one-shot prompt without opening the side panel and explaining what to do

The two are not mutually exclusive. Many writers keep Gemini side panel open for document-wide context tasks ("rewrite the whole document in a friendlier tone") and use Prompt Anything Pro for snippet-level prompts ("rewrite this paragraph from passive to active").

Google Docs is a contenteditable surface, which means Chrome extensions can read selected text but cannot directly modify the document via the DOM. Prompt Anything Pro works with that constraint:

1. Select text the normal way — drag, double-click for a word, triple-click for a paragraph. Ctrl/Cmd+A selects everything in the active doc body.
2. Trigger Prompt Anything Pro — either click the floating action button that appears near the selection, or press your configured keyboard shortcut (default: Cmd+Shift+P on Mac, Ctrl+Shift+P on Windows; the extension's settings page lets you remap if Docs has a conflict).
3. The popup opens with the selected text in the input — you don't have to paste. Choose a saved prompt from your library or type a one-off.
4. Review the AI response in the popup, then click Copy.
5. Replace or insert — to replace the original selection: don't deselect, just Cmd+Shift+V (Mac) or Ctrl+Shift+V (Windows) to paste-without-formatting. Google Docs replaces the selected text with the pasted content, preserving the doc's current style.

One catch: if you click in the popup and then click back into the doc, your original selection is lost. Train yourself to use Cmd/Ctrl+Shift+V from the popup-still-open state, or use Suggesting mode (next tip).

The classic Google Docs use case Gemini does well but Prompt Anything Pro does cheaper-per-call via BYOK. The workflow:

1. For docs under 75,000 words (roughly 100K tokens, the common context window in 2026): select everything (Cmd/Ctrl+A), trigger Prompt Anything Pro, use this prompt verbatim: "Write an executive summary as exactly 5 bullets. Each bullet is one finding or recommendation. No bullet exceeds 25 words. No preamble — return only the bullets."
2. For longer docs: split into sections of 3,000–5,000 words at H1 boundaries, summarise each section, then run a final consolidating prompt on the section summaries: "These are summaries of consecutive sections of one document. Consolidate into 5 bullets representing the document's top findings. No preamble."
3. Paste the result at the top of the doc as a styled callout (Insert → Building blocks → Highlight) so the reader sees it before the body.

The "no preamble" instruction is non-negotiable. Without it, most models add "Here is the executive summary:" which you'd then manually delete. Saying it once per prompt saves about 8 seconds per pass.

Generic "rewrite this for X audience" prompts produce mush. Prompts that work in Google Docs — copy these:

• Technical → Executive (board-deck-ready): "Rewrite this for a non-technical board audience. Keep the same 3 facts. Remove every TLA and acronym. Add a one-line plain-English consequence for the business. No fluff additions."
• Formal → Blog-post casual: "Rewrite this in a conversational blog post tone. Use 'you' to address the reader. Keep the same claims and numbers. No emojis. No exclamation marks."
• Long-paragraph → Skimmable: "Restructure this single paragraph into a short intro sentence + a 3-item bulleted list + a one-sentence conclusion. Don't add information."
• Passive → Active voice: "Rewrite every passive-voice sentence in active voice. Don't change sentences that are already active. Don't add or remove sentences. Return only the rewritten text."
• Heading inflation: Select the H2 headings only, then prompt: "Improve each heading to be specific and benefit-focused. Limit to 8 words. Return the headings as a numbered list in the same order as the input. No preamble."

Save each one to Prompt Anything Pro's prompt library so they're a click away across all docs.

Things that will surprise you the first time:

• The extension can't see your doc. It reads only what you select. "Summarise this whole document" without selecting anything first will summarise an empty string. Always Cmd/Ctrl+A first.
• Headers and footers are separate selection surfaces. If you need to apply a prompt to a footer, click into the footer first, then select the footer text — selecting body text and clicking into the footer to paste doesn't work.
• Pasting overwrites footnotes if they're in your selection. Google Docs footnote markers are part of the paragraph DOM. Paste-replace removes them. Use Suggesting mode for any doc with footnotes.
• Tracked-changes mode is sticky. If the doc is in Suggesting mode, every Prompt Anything Pro paste is suggested, not direct. Switch back to Editing (Cmd/Ctrl+Alt+Shift+Z) when you're done.
• Image alt text isn't selectable from the body. To rewrite image alt text, right-click the image → Image options → Alt text — and copy the existing alt text into Prompt Anything Pro's input manually.
• Very long selections (>50K words) may exceed the AI model's input context. The extension will return an error from the AI provider; split the selection.

For confidential documents: Prompt Anything Pro sends only the text you select to the model provider you've configured. It doesn't send anything else from the doc, your Google account, or Drive. The privacy boundary is exactly the text you highlight.