AI-assisted workflows

From Bilingual Draft to Website Page: The Handoff Rules I Clarified

The editorial inputs I clarified for my personal site, from page text and language pairing to image responsibilities and publication status.

Huiyang Xie

On this page
    1. Separate the article’s voice from its page metadata
    2. Make the opening answer and final FAQ part of the deliverable
    3. Pair the articles explicitly, then check the destinations
    4. Keep dates and permission out of the guesses
    5. Give images a producer, not just a mention
    6. The checklist I want before page implementation
    7. What I wanted the handoff to achieve
  1. FAQ

The review of my bilingual website left me with a practical question: what exactly should the person building the page receive?

The writing process already covered much of the work. It had editorial review, SEO considerations, language adaptation and a website handoff. The website also had its own templates and content fields. The remaining difficulty was that broad instructions did not always identify the precise content to deliver, its position, or who would confirm the unfinished parts.

Some decisions needed to be made explicit. Others were already sound and simply needed to survive the handoff. I wanted a package that made those differences visible before page implementation began.

Separate the article’s voice from its page metadata

A manuscript gives the implementer a headline and prose. A website may need several additional pieces of text, each with a different job.

For my site, I made the handoff explicit about the article title, the short page description, the SEO title and the SEO description. The SEO fields already existed in the site’s content structure. The change was to make their delivery and checks clear in both languages.

That matters because writing a short summary is an editorial decision. An implementer should not have to decide which claim to emphasize by extracting a sentence from the introduction. The extracted sentence might make sense in context but promise too much when it stands alone.

I review each field against the article:

  • Does the title accurately describe the subject?
  • Does the page summary explain the article without introducing a new result?
  • Does the SEO title state a useful, recognizable topic?
  • Does the SEO description describe what the reader will actually find?

English and Chinese need their own wording and length checks. A Chinese description should not be stretched to follow the shape of the English sentence, and a shorter field should not lose a qualification that makes a claim accurate.

The checks belong to my site’s editorial requirements. Passing them establishes that the supplied fields fit those requirements; it does not establish a ranking benefit or guarantee how a search service will display the page.

Make the opening answer and final FAQ part of the deliverable

I explicitly confirmed two requirements: every article must have a Quick Answer at the very beginning of its body, and a relevant FAQ at the end.

The opening answer should answer the article’s main question in its first sentence. It should work for a reader who wants an immediate orientation before deciding whether to continue.

The FAQ has a different job. It addresses related uncertainties that remain after the main explanation. For this article, questions about language pairing and publication permission are useful. A generic question added merely to fill a field would not improve the handoff.

The position rule also removed a source of ambiguity. The website already placed the Quick Answer before the ordinary body text. I was clarifying the writing requirement to match that arrangement, not asking for a new layout.

I wanted one approved source for each block. If the website renders Quick Answer and FAQ from structured fields, the implementer should not paste another version into the article body. The same principle applies when the visible FAQ feeds the site’s structured-data helper: the answers should not be rewritten into a separate version for that purpose.

This leaves a straightforward division. Editorial delivery supplies the actual words. Website implementation preserves them and checks that they appear once, in the intended place.

Pair the articles explicitly, then check the destinations

Two files being “the English one” and “the Chinese one” is not enough information for a website.

I confirmed that new article pairs on my site would use the same concise English slug, with language represented by the route, and the same translationKey to identify the pair. This brought conflicting guidance back into one agreed rule.

It did not authorize changing every existing Chinese URL. Published articles with earlier URL conventions remained separate from the rules for new work.

The useful result is a clear identity for each pair. The implementer can see which English and Chinese articles belong together without interpreting their translated titles. I regard this as a maintainability decision, not a claim that an English slug will improve Chinese search rankings.

Related links need their own check. A shared slug for a new pair does not mean every older article has identical slugs in both languages.

That was already a real issue on my site: some Chinese article relationships had used an English slug where the Chinese destination used a different one. A later review confirmed that the fixes were already present. The handoff still needed to require the writer to check each selected destination.

For every related article, I want to know:

  • Which actual page is being recommended?
  • Is it available in the current article’s language?
  • Is that version published and accessible?
  • Does the relationship help this reader?

The last question prevents a technically valid link from becoming an arbitrary card. The publication check prevents the opposite problem: recommending a useful page that visitors cannot yet open.

If a destination is still a draft, I leave out the live link or keep a safe plain-text reference where it adds value. I do not change the destination’s publication state merely to complete a related-content section.

Keep dates and permission out of the guesses

A writing deadline, a review date and a publication date can all be different.

The handoff preserves that distinction. An unpublished article should not acquire a publication date because the writer finished it today or because someone hopes to release it tomorrow. In my site’s process, the release operator fills the actual publishDate when publication is authorized and takes place.

This was an existing principle worth keeping explicit. It did not need to be reinvented as a new feature.

Approval also needs a clear object. When I approve the article, I am approving the content I reviewed. That does not prove that an unseen website conversion preserves it, that the sharing images exist, or that the page is ready to release.

The next operator therefore needs the exact approved content, its language pair and its current status. If implementation exposes missing copy, that question returns to editorial review. It should not be answered by silently adding a sentence that the owner never saw.

These details can feel administrative until a handoff depends on them. Then they determine whether the next person is implementing a decision or making a new one.

Give images a producer, not just a mention

Images were an important place to clarify ownership.

For article sharing images, my website already had an OG template. The writer’s responsibility is to supply the localized title, summary, category and article identity that the template needs. The website stage is responsible for producing the actual image files and checking their integration.

On my site, that means a static PNG for each language using the existing template. Delivering the text fields is a completed editorial input; the corresponding image can still be pending.

I want that pending status to be visible. “The template exists” is not a useful substitute for saying whether this article’s image has been generated. At the same time, an unfinished image does not require the writer to design a second visual system.

Body images are a separate decision. If an article needs one, the handoff should identify the intended image or source, its place in the story, the caption and alternative text, and the relevant permission information. If it has not been produced, the package should say who owns that work.

An article that does not need a body image should not acquire a decorative placeholder just to make the package look fuller. For a text-led explanation, a clear structure may serve the reader better.

The checklist I want before page implementation

The practical handoff can remain compact. I want the next person to receive these six groups of information:

  1. Reviewed content and page text. The English and Chinese manuscripts, plus separately considered article titles, summaries and SEO fields.
  2. Required answer blocks. The exact Quick Answer and FAQ text, with their intended positions and a clear single source.
  3. Article identity and relationships. The shared pairing information, locale-specific routes and checked related destinations.
  4. Status and dates. Whether content has been approved, what remains pending, and no invented publication or modification dates.
  5. Image inputs and responsibility. What text or assets are available, what still needs production, and who will produce and check them.
  6. The next permitted action. What the website operator may implement, what needs another editorial decision, and whether publication has been separately authorized.

This checklist does not replace the website’s tests. It makes the input to those tests clearer.

The website stage still has to check conversion, layout, links, metadata, images and the behaviour of unpublished content. Those checks concern the implemented page. A complete editorial package cannot perform them in advance.

What I wanted the handoff to achieve

The point of clarifying these rules was to reduce the editorial questions left hidden inside implementation.

That does not require the writer to own every technical step. It requires the writer to deliver the decisions it owns, and to name the work that remains elsewhere.

In this review, I brought several kinds of work together: explicit field delivery, reaffirmed page positions, consistent language identity, checked destinations and preserved release boundaries. Some were additions to the handoff. Others were confirmations of rules or capabilities already in place.

That is the standard I would use again: the next person should be able to build from the package without guessing the article’s meaning or assuming a permission that has not been given.

For a broader discussion of where one stage ends and the next begins, see How to Define “Done” in an AI-Assisted Workflow.

Questions

Frequently asked questions

Why are two approved manuscripts not enough for a bilingual website handoff?

The website also needs decisions about page and SEO text, answer blocks, language pairing, links, images and publication status. If those inputs are absent, the implementer may have to make editorial choices that were never reviewed.

Does every website need Quick Answer and FAQ in these exact positions?

These are requirements I confirmed for my own site, not a universal layout rule. The broader lesson is to make a site’s required content and placement explicit in the handoff, then check that implementation preserves them.

If both languages share a slug, do related links still need language checks?

Yes. Existing destinations may follow different URL conventions, and availability can differ by language. Check the actual published target for each locale rather than inferring it from the new article’s pairing rule.

Can content be handed over while an OG image is still pending?

Yes, provided the supplied text is complete and the pending image and responsible producer are explicit. That is an editorial handoff with remaining website work, not evidence that the article is ready for publication.

Does approving the article authorize publication?

No. Content approval covers the reviewed text. Website implementation and its checks still need to happen, and publication needs its own explicit authorization.

Share

Found this useful? Share it.

LinkedInXCopy link

Discuss

Want to discuss the idea?

Discuss with me

Author

About the author

Huiyang Xie is a marketing professional based in Greater Vancouver, Canada, working across digital marketing, content, websites, AI-assisted workflows, and cross-cultural marketing. Her work explores how AI can support practical marketing processes while keeping factual review, brand judgment, and human approval in the loop.

Continue reading

Related work and reading

Blog article

How to Define “Done” in an AI-Assisted Workflow

Open