Google Docs: structured brief and exports

What happened

An agent created a synthetic project brief through native Google Docs requests, then inspected its semantic structure and actual DOCX/PDF exports. The six named title/headings, three-item ordered list, five-item bullet list, four-row budget table, Café / résumé text, reference hyperlink and three native date chips were present. Independent export parsing preserved the required facts; the PDF rendered cleanly as one page. Invalid initial API enum values and formatting/index corrections were evaluator authoring errors, corrected before final checks. This is connector-assisted construction and export evidence, with no Gemini or collaboration claim.

Assertions and observed results

• PassedNative heading structure

  Expected: The title plus Context, Scope, Deliverables, Acceptance checks and Timeline have native heading styles.

  Observed: Native readback preserved all six named title/headings; Deliverables used Heading 2 and the other sections Heading 1.

• PassedNative ordered and unordered lists

  Expected: Separate native lists contain three scope items and five acceptance items.

  Observed: Native list identifiers showed two lists containing three and five paragraphs respectively.

• PassedNative date chips

  Expected: Three native ISO date chips retain 12, 19 and 23 October 2026.

  Observed: Native date-element readback found all three expected dates.

• PassedDOCX headings and budget table

  Expected: The DOCX export preserves six title/headings and the four-row, three-column table.

  Observed: Independent DOCX parsing found every expected heading and all table cells, including Total: 5 and Budget GBP 1000.00.

• PassedDOCX list semantics and reference link

  Expected: The export retains three numbered items, five bullet items and a clickable reference link.

  Observed: DOCX numbering definitions contained three decimal and five bullet paragraphs. The hyperlink relationship targeted https://example.com/brief.

• PassedPDF and DOCX text round trip

  Expected: Both exports retain the required headings, Unicode, GBP 1000.00 and three dates.

  Observed: Both formats retained Café / résumé, all six title/headings, the budget and dates. The PDF had one page and passed visual inspection.

• PassedPublication export metadata

  Expected: Publication exports exclude account identifiers and private cloud links.

  Observed: A scan of the listed native exports found no account name/email, private file IDs or private cloud URLs. Raw UI screenshots and connector receipts were excluded.

Reproduce the task

1. Create a new scratch Google Doc from project-brief.md and expected facts. Do not change sharing or existing documents.
2. Apply native Title, Heading 1 and Heading 2 styles to the six named sections. Create a three-item ordered list and five-item unordered list.
3. Insert a four-row, three-column table containing Item, Count and Budget GBP, the two deliverables and the GBP 1000.00 total.
4. Apply an ordinary clickable hyperlink to https://example.com/brief and insert native ISO-formatted date chips for 12, 19 and 23 October 2026.
5. Read the document's native structure back, confirm list item counts, headings and date elements, and inspect the browser document.
6. Export through Google Docs as DOCX and PDF. Independently check DOCX headings, numbering, bullet lists, table and hyperlink relationships. Check required text and dates in both exports and visually inspect the one-page PDF.

Inputs and saved outputs

The downloads contain synthetic test material. Account identifiers and private project links are excluded. Hashes identify the exact published bytes.

• Synthetic brief fixture ↓SHA256 2072a81e5b18619c79cdc5da3b181b795656bc9d0371e438ea4647f1e30fde78
• Expected brief structure ↓SHA256 05b16d22ad6372a8f83237be324d2296bc4b070b34e634e5e841e4760c6d2fa9
• Native Google Docs DOCX export ↓SHA256 984697e08d7f2b971ee3a565dced3df0bb733758460d4bab26596bee977e4e53
• Native Google Docs PDF export ↓SHA256 9f598dd3c8731fd408a1fb4a3c8ccc22591df823b78f991991a8fa67f5d2950b
• Rendered native PDF page ↓SHA256 28a141b7c275d649f38484a41cec9ee90d45b64a9aecd2c0327d2d3ddc3bbfe9

Limits of this result

• One synthetic task in one connected account on one date. This establishes only the listed observed results; it is not a general reliability, speed, security, usability, accessibility or customer-support evaluation.
• No peer product ran the same task in this session, so this is not comparative evidence and does not establish a winner.
• The fixture and intermediate adapter work were supplied by the evaluator. No product AI quality was evaluated.
• Construction and formatting were connector-assisted. This does not evaluate manual typing/formatting friction or a first-time user's experience.
• Gemini, multi-user collaboration, suggestions, comments, offline editing, complex page layouts and imported-document compatibility were not tested.
• Native date chips were verified in Google Docs. Their dates were verified in exported text, but editable smart-chip behaviour in Microsoft Word was not tested.
• The initial authoring request used two invalid API enum values and was rejected without document changes. The evaluator corrected those values, one date insertion position, table text sizing and an inherited empty heading before final verification. These were evaluator request/formatting corrections, not failed product assertions.
• PDF text extraction required whitespace normalization, and its reading order places the three date labels apart from their values. Visual inspection confirms the one-page timeline is in the intended order.

This record does not change a general product score or name an overall winner. Changes to editorial opinions or recommendations require review.

Continue the research