Review AI output and share web pages
Review pending placeholders, finished canvas results, AI badges, permissions, source context, and webpage-style sharing before using AI output.
AI output needs review before it becomes shared work
AI Studio is built to put generated work back on the canvas: a draft note, a table, an image, an audio result, or a webpage-style preview. That is useful because the result lands beside the work that shaped it. It also means an unfinished or unreviewed draft can become part of shared work faster than you intended.
Preview the result before you present it, share it, export it, assign work from it, or ask a teammate to act on it. Check the source, facts, format, permissions, and audience. AI can sound finished while it is still wrong.
Only webpage-style AI output has its own shareable page flow. Docs, tables, images, audio, and source objects stay on the canvas as canvas objects. To show those to someone else, share the canvas with the right permission, export or download where the object supports it, or copy the reviewed content into the place where it belongs.
For generation, read Use AI Studio. For recovery, read When an AI result is not right.
Availability and permissions
| Item | Details |
|---|---|
| Available on | Workspaces with AI Studio enabled and usable workspace AI credits. |
| Available for | Web app and desktop app canvas editor. |
| Where generation controls appear | Writable, non-embedded canvas contexts after the editor is ready. |
| Where generation controls are not available | Read-only canvases, comment-only access, embedded views, unsupported object states, and credit-blocked workspaces. |
| Who can preview a result | People who can open the finished canvas object or generated result. |
| Who can reuse a result | People with the canvas or item permission needed for that action. |
| What can be shared as a generated page | Webpage-style AI output. |
| What cannot be shared directly from AI Studio | Generated docs, tables, images, audio, source objects, and Object Chat answers. Share the canvas, export/download where supported, or copy reviewed content instead. |
How generated output appears
When ALLO accepts an AI Studio request, ALLO creates the placeholder on the canvas, applies progress updates, and writes the final result. The canvas object is the source of truth. Current AI Studio output does not require a manual drag or insert step from a chat card.
| State | What it means |
|---|---|
| Pending placeholder | AI work is still running. You can select, resize, or delete the placeholder. Preview, edit, download, copy, duplicate, lock, and object-specific actions stay blocked until output is ready. |
| Finished generated object | The result has been applied to the canvas. Review it like any other work item before you reuse or share it. |
| AI badge on a finished object | AI Studio results and AI-edited objects show an AI badge. Use the badge to ask follow-up questions about that object. The badge is separate from human comments. |
| Webpage-style result | The result opens as a live preview. Review the page, links, title, description, and canvas thumbnail before using the webpage share flow. |
| Legacy chat card | Older AI chat history can contain card-shaped responses. For current AI Studio generation, rely on the canvas object and its placeholder state. |
For webpage-style output, the preview is part of the review surface. Check the visible card, title, and opening state before you use it in a handoff or share it as a standalone result.
Review each output type
Different outputs fail in different ways. The review step should match the thing AI created.
| Output type | What to check |
|---|---|
| Document or text-style output | Facts, links, citations, tone, audience, and whether the draft says more than the source proves. |
| Table or structured output | Columns, units, row counts, calculations, missing fields, and whether every source file finished processing. |
| Image | Text rendering, small details, brand fit, distortions, sensitive content, and whether the image is safe to show outside the team. |
| Audio | Playback, pacing, names, claims, source fit, and whether the audio says anything that needs human approval. |
| Webpage-style output | Layout, links, copy, responsiveness, title, description, thumbnail, and whether the page makes unsupported claims. |
| Object-level AI answer | Whether the answer refers to the right canvas object. Turn useful parts into canvas content, a comment, a task, or a new AI Studio request only after review. |
Handle pending placeholders deliberately
A pending placeholder is not a broken result. It is the visible sign that generation is still running. File-heavy prompts, image generation, audio, and webpage-style output take longer than short text drafts.
Do not submit the same expensive request again just because the placeholder is still processing. Refresh the canvas first. Then check whether the placeholder still exists, whether an attached file failed, and whether the workspace ran into an AI credit block.
Delete a pending placeholder only when you mean to cancel or abandon that output. Deleting the processing placeholder, or deleting the canvas that contains it, cancels the generation. After deletion, the result is no longer something to recover from the canvas.
Review checklist
| Check | Why it matters |
|---|---|
| Source | Confirm AI used the intended canvas objects, files, links, or prompt. |
| Accuracy | Verify names, dates, numbers, claims, decisions, and references. |
| Format | Check whether the output is the requested document, table, image, audio, or webpage-style result. |
| Tone | Make sure the result fits the audience, especially for clients or executives. |
| Completeness | Look for missing constraints, skipped sections, or unanswered questions. |
| Sensitive content | Remove private, internal, legal, billing, or customer data before sharing externally. |
| Permissions | Confirm the people you share with should see the generated content and its source context. |
| Cost sense | Edit a close result manually instead of spending more credits on repeated regeneration. |
Reuse output on the canvas
If the output is useful, keep it near the source material or leave it where AI Studio created it. Add comments for human review, mention the owner, or use Team chat for live discussion.
Use the finished object as canvas content only after review. Edit it manually when the result is close. Use the AI badge or item-level AI action when you want to ask about the generated object itself. Use AI Studio again when you need a new object, a different format, or a cleaner generation from better source context.
If the output is not useful, delete it or move it out of the main review area. Do not leave misleading AI drafts beside final work without labeling or cleanup.
Share webpage-style output safely
AI output inherits the audience of the place where you put it. If generated output sits on a shared canvas, people who can open that canvas can see the output and the surrounding context.
For direct AI output sharing, use the webpage-style output flow. That is the AI Studio result type designed to open as a previewable page and then be shared as a page. Generated images, docs, tables, audio, and Object Chat answers do not have the same direct share flow.
Before sharing, open the generated page yourself and make sure the version is the one you intend to send. If the canvas still contains older attempts, rename or move them so reviewers do not open the wrong artifact.
Before sharing externally, check for internal names, hidden assumptions, private files, customer data, pricing claims, legal language, and unsupported statements. AI can sound confident even when it is wrong.
Use Share work with teammates for access choices and Work with comments and mentions for review loops.
Examples
If AI creates a customer-facing summary from internal notes, preview it for internal-only language, unsupported promises, and missing context. Rewrite before sharing with the client.
If AI creates a table from several files, verify columns, units, row counts, and whether any file failed to upload. A clean-looking table can still be based on incomplete source material.
If AI creates an image, inspect text and small details. AI images can contain distorted text or visual artifacts.
If AI creates webpage-style output, preview layout and links before sending it to stakeholders.
Common mistakes
Do not treat preview as approval. Preview is where you review; approval still belongs to the responsible person.
Do not share the canvas broadly just because the AI output looks useful. Check the source material and audience.
Do not look for a direct share action on generated docs, images, tables, audio, or Object Chat answers. If the output is not a webpage-style result, share the canvas or use the object's normal export, download, copy, or review path.
Do not regenerate endlessly when the result is close. Edit manually, ask about the generated object, or clarify the prompt once.
Do not ignore pending or failed context. A failed upload is not source context for the request.
Do not mistake the AI badge for a comment. AI badges open object-level AI context. Human review still belongs in comments, Team chat, or a meeting.
When preview or sharing fails
Preview is missing. The output is still pending, failed, deleted, or not a previewable result. Refresh the canvas before submitting the same request again.
The AI badge is missing. Pending uploads and pending generation do not show the AI badge. Manual or uploaded objects with no AI-created or AI-edited history do not show one either.
A generated webpage opens in preview but the canvas card still looks generic. Refresh the canvas. If the card still has the wrong title, description, or thumbnail after the preview works, contact support with the canvas link, approximate time, and output type.
A teammate cannot open the result. Check the canvas or item permission in Share work with teammates. Starting an AI chat about an object does not create a human comment thread or change who can open the object.
A finished result is wrong. Use When an AI result is not right. Revise the prompt, attach better source context, ask about the generated object, edit manually, or generate a new version.
Related articles
- Use AI Studio in a canvas
- Give feedback with comments and mentions
- Share work with teammates
- Use AI Studio
- Ask about a canvas item
- When an AI result is not right