Content workflow

Tailwin’s content workflow starts with a measured visibility gap and ends with evidence about publication, crawler discovery, and citations.

1. Select a real gap

Open Content and choose a candidate tied to an active query. Before drafting, inspect the answer engines’ current sources and confirm that a new or improved page is the right fix. Access, factual, or profile problems should be corrected directly instead of hidden inside a new article.

2. Establish the brand voice

Generate or refresh the site voice profile from approved public writing. Review the profile before relying on it. The profile describes writing patterns; it does not authorize Tailwin to invent opinions or claims.

3. Generate and review

Tailwin grounds the draft in the business profile and selected gap, applies the voice profile, then runs deterministic quality checks for readability, SEO, citability, length, structure, banned language, and honesty.

An operator remains responsible for:

• factual accuracy and source rights;
• regulated or performance claims;
• prices, dates, names, and legal statements;
• customer-specific promises;
• final editorial judgment.

Edit the draft in the review screen and inspect the visible pass/fail checks. A passing automated gate is not legal or subject-matter approval.

Start from one competitor page

Owners and admins can use Competitor page to original draft in Content. Select your target site and enter one public HTTPS page URL without query parameters. Tailwin checks that site's robots policy, then reads only the supplied page. It does not follow redirects, crawl links, render JavaScript, or reproduce the complete browser layout. Use the final page URL if the source redirects.

Confirm the keyword analysis separately. The interface asks you to authorize up to $0.02 for a new request, using the reviewed DataForSEO Labs rate for up to 50 rows. The interface uses United States/English estimates. It shows the provider's recorded cost, date, location, and language with the result. Keyword positions, volume, and traffic are vendor estimates, not measured site traffic or verified current rankings. An empty result means the provider returned no estimates for that page.

The analysis shows the extracted title, headings, word count, declared publication date, and entities explicitly named in page metadata. Missing dates and entities remain missing. Extraction reflects static HTML; content loaded by JavaScript may be absent. Responses larger than 512 KB are rejected. Long extracted text can be shown as truncated, and that analysis cannot be used to draft because the complete source is needed for comparison.

An analysis is cached for 12 hours for the same workspace, target site, page, location, and language. A valid cache hit makes no new research request. Each workspace can start at most 20 new analyses per day, including attempts that fail. Pending requests and uncertain provider charges remain blocked from automatic retry until their receipts are reviewed.

After reviewing the evidence, choose an available draft model and separately confirm the AI charge and content credit. Language, locale, market, tone, and depth come from your shared chat preferences. This action creates an article for the selected site; it does not promise a ranking improvement. Workspace model restrictions and AI spending caps still apply.

Own-brand mentions are off by default. If you allow them and the generated text mentions a configured brand alias or domain, Tailwin adds a publisher relationship disclosure. Keep the site's brand aliases accurate. Similarity and brand checks are deterministic heuristics: they can miss semantic copying or unfamiliar brand variants, so review the source, originality, claims, and disclosure yourself. A blocked draft is not saved as editable copied text; incurred AI usage remains recorded.

A successful real draft opens in Content for editorial review with no approval or publication applied. This workflow does not send approval emails automatically. In demo mode, drafting returns a clearly labelled simulated preview and creates no publishable Content piece. Demo AI does not make the separate keyword-analysis request free.

4. Approve

Scheduled drafts require agency approval and can optionally require client approval. The approver receives a tokenized, branded page and can approve or reject without receiving access to unrelated workspace data.

Do not forward an approval link beyond the intended reviewer. Revoke or regenerate access if it is exposed.

5. Publish

Available destinations include WordPress, Shopify, Ghost, dev.to, Hashnode, Blogger, LinkedIn, Facebook, export formats, and a signed webhook path where configured. Credentials are stored encrypted and scoped to the organization.

Publishing and approval are separate states. A successful approval does not prove the CMS accepted the post; review the publication result and URL.

6. Verify the outcome

The content receipt can show:

1. publication time and URL;
2. first observed crawler fetch by bot;
3. first observed citation by engine;
4. movement for the target query.

The receipt is partner-branded, tokenized, revocable, and noindexed. Missing milestones render as unmeasured. Tailwin also freezes content outcomes before raw crawler-hit retention expires so long-term evidence can survive the lower-level event window.

Signed provenance

When configured, Tailwin creates an Ed25519 content credential for a published piece and exposes a public verification endpoint. The credential proves what Tailwin signed; it does not prove every factual claim in the content. A signing failure does not block publishing and must remain visible as a missing credential.

Safe operating checklist

• The query reflects buyer intent.
• Existing sources and current answer reviewed.
• Business profile and voice profile are current.
• Claims, figures, names, and links verified.
• Quality-check failures resolved or consciously accepted.
• Correct approver selected.
• Correct destination and publication time confirmed.
• Published URL opens without authentication.
• Receipt milestones interpreted as observations, not guarantees.