For Staff: Help Articles, AI Drafts, and Website Artwork in Circle Read

Who this guide is for

This article is for staff and hosts who operate a Circle Read deployment: librarians, instructional technologists, community managers, and volunteer admins. Members rarely need this page; it explains how your team scales help content and public visuals without living in spreadsheets.

Editorial strategy in 2026

Organizations that want steady discovery often plan ten to fifteen substantive articles per month across onboarding, SEO topics (book clubs, study groups, faith formation), and feature education. Quality beats noise: long-form guides that answer real search questions outperform thin keyword pages.

Batch thematically (“Back to campus,” “Advent,” “Audiobook clubs”) so marketing visuals and articles reinforce each other.

Help articles: from idea to publish

In the staff area, open Help articles (sometimes labeled alongside other content tools).

Title and outline

• Use suggested title chips for common patterns (getting started, hosts, students, faith groups).
• Add outline starters so writers and models share the same skeleton.
• Paste institution-specific notes (policies, links, contact paths) into the outline field—the model should not invent those.

One-action drafting

Write first draft with assistant sends your title + outline to the configured LLM and returns Markdown you can edit. The app is tuned for long output (thousands of words when appropriate): adjust your provider settings if you need even larger completions.

Always fact-check and localize drafts. Models summarize product patterns; they do not know your unpublished roadmap or legal constraints.

Review checklist before publish

• Links and contact info are real.
• Features described match what your tenant actually ships.
• Tone matches your community (school vs. church vs. public library).
• Headings use ## / ### so the rendered page is scannable.

Hero artwork

When your workflow supports it, attach a hero image to an article so list pages and social previews look intentional. Many teams generate article header artwork under Website visuals, then associate the asset in the article editor.

Website visuals: heroes, banners, sections

Under Website visuals, staff generate images or short videos for defined placements:

• Homepage hero — Behind the main headline on your marketing home.
• Banner strip — Thin band under the top bar for campaigns.
• Section backgrounds — Subtle art behind feature grids.
• Article / help headers — Illustrations for long reads.

Outputs are saved to your project storage (not only the vendor’s short-lived preview). Placement controls let you mark what is live vs. hidden so you can stage seasonal art safely.

Use prompt suggestion chips for literary, inclusive, and loop-friendly video descriptions—then edit every prompt before generating.

LLM limits and “heavy content”

Server defaults favor large completion budgets for help drafts and the member reading assistant so answers and articles are not artificially truncated. Operators can raise or lower the cap with an environment variable (see project .env.example: SHORTAPI_LLM_MAX_OUTPUT_TOKENS). Provider maximums still apply—if a model rejects a huge request, reduce the cap or split the task.

The reading assistant accepts long pasted excerpts for close reading; extremely large pastes may still hit upstream context windows—trim if you see errors.

Cadence that teams actually keep

• Weekly: one help fix or FAQ from support tickets.
• Monthly: one refreshed hero or banner + three to five new articles.
• Quarterly: audit published articles for outdated screenshots or feature names.

Accessibility and brand

Prefer descriptive alt text on heroes, sufficient contrast on text-over-image heroes, and captions on video loops when autoplay is used. Your brand guidelines should cover typography on marketing pages separately from in-app UI.

Governance

Decide who may publish (single comms lead vs. any platform admin). Keep a style guide link in your staff channel. Log major changes in your internal changelog so facilitators are not surprised.

Closing

Circle Read separates member simplicity from staff power: members get calm reading tools; staff get LLM-assisted drafting and visual generation with storage and placement discipline. Used together, you can run a 2026 content program that is both search-visible and true to what your community actually offers.