AI Content Research & Generation

Turn a scraped source page into a grounded, brand-aware blog draft in under a minute.

TL;DR

• What it is — the pipeline that produces your first draft. Content Studio gathers live research on the topic, then generates the blog in your voice using AI.
• Who it's for — marketers who want a grounded, on-brand draft without prompt engineering, and admins who want to pick the model per run for flagship content.
• Top outcome — from "I have a clean source page and a brand profile" to "a draft is open in the editor" in about 30 to 60 seconds.

At a glance

How to find it

Generation starts from the Sources side of Content Studio, and it always ends in the editor on the resulting draft.

1. Sources → pick a domain → pick a source page → Suggest new blogs → pick a suggestion → Generate.
2. Sources → pick a source page → Recreate from this source.

Screenshot [SCREENSHOT: aigen-source-actions.png — source detail page with "Recreate" and "Suggest new" buttons outlined]

What is AI Content Generation?

The problem it solves

Off-the-shelf AI writing tools have three consistent failure modes:

1. No grounding. The AI writes whatever it "knows" about your topic, which is usually two years stale and wrong about your industry's specifics.
2. No brand voice. Without a saved profile, every blog sounds like "ChatGPT wrote this." You end up rebuilding prompts by hand every time.
3. Loose output. Free-form generation means someone downstream has to extract a title, a slug, a summary, and metadata from a blob of prose.

Content Studio solves all three. Grounding: research runs on the live web first, and the generator only sees your clean source page plus that research. Brand voice: every generation reads your brand profile — company name, description, industries, products, customers, tone, style, and call-to-action. Structure: the AI returns a complete, well-formed draft with title, slug, body, summary, meta description, keywords, and category. No parsing required.

What you get

• A draft that's already 70–80% publishable for typical hospitality and SaaS use cases.
• Consistent structure. Every draft has a title, a slug, body, summary, meta description, keywords, tags, category, and estimated reading time.
• Per-run tone and style overrides. Need a particular post to be conversational rather than your default "professional"? Override for that one run without changing your brand profile.
• Per-run model override. Swap in a premium model for your flagship posts while keeping a fast default for bulk drafting.
• Source-grounded. The generator never starts from a blank page — it always works from a scraped source's clean text.

How it's different

• Research runs first, writing second. Most AI writers try to browse and compose in a single step, which degrades both. Content Studio does them in two explicit stages.
• Brand profile on every run. Voice, tone, audience, products, and CTA flow into every draft automatically.
• Dedup-aware. The generator sees the slugs of your existing blogs and avoids duplicating them.
• Cheap default, premium upgrade. Use the fast model for daily work; switch to the premium model for flagship content with one setting.

Customer scenarios

1. "Build a 20-blog library from our own pages." For each page on your sitemap: scrape → generate (Recreate) → edit → publish. One week, one person, no writing agency.
2. "We have a single competitor listing we want to out-rank." Scrape their listing → Suggest new blogs → pick the angle that fits your positioning → generate → refine → publish.
3. "Rank for a specific keyword set." Use SEO-targeted suggestions with your target keywords → pick the best match → generate → publish. SEO-tuned titles and meta descriptions are baked in.
4. "Switch our premium brand to a stronger model." Update the brand profile's default model once. Every future generation uses the new model; existing drafts are unaffected.

How it fits with the rest of Lodgestory

• Upstream: Sources & Scraping produces the clean source pages this module consumes.
• Config: your brand profile (voice, tone, audience, call-to-action) lives in Content Studio settings. Every generation reads it.
• Downstream: every successful generation opens the Editor on a Draft blog.

Screenshot [SCREENSHOT: aigen-landing.png — source detail with Generate button, config dropdown visible]

Core concepts

Quick Start — three shapes of generation

Content Studio gives you three ways to start from a source:

A. Recreate a blog from a source (fastest path)

Open the source page and click Recreate from this source. Content Studio reshapes the source into a blog in your voice, respecting your brand profile and avoiding slugs you've already published.

Use when: you own the source page and want a fresh, on-brand version of it.

B. Get suggestions first, then generate

Open the source page and click Suggest new blogs. Content Studio proposes five angles, each with a title, a short description, target keywords, and a search-volume hint (High / Medium / Low). Pick the one that fits and click Generate.

Use when: you want to fan out from a content-rich source into multiple distinct blogs.

C. SEO-targeted suggestions

Open the source page and click SEO-targeted suggestions. Add three to five target keywords. Content Studio proposes angles most likely to rank for that keyword set.

Use when: you're targeting a specific keyword cluster and need titles + meta descriptions tuned for it.

Screenshot [SCREENSHOT: aigen-suggestions.png — list of 5 suggestion cards with title, description, target keywords, search-volume hint]

How it works

1. You pick a source page (your own content, a competitor's listing, or any page you've scraped).
2. Content Studio reads your brand profile — voice, tone, style, products, audience, CTA — and composes a system instruction that anchors the generation in your brand.
3. A live-web research pass gathers current facts about the topic. The research isn't quoted in the final blog; it's used to keep the writing grounded in today's reality instead of stale training data.
4. The generator writes the blog in your voice, using the source page and the research as reference. It sees your existing slugs to avoid duplicating them.
5. The output lands as a Draft. The editor opens automatically on the new blog.

You can override tone, style, or model for a specific run if needed — useful for A/B testing or for flagship content that deserves a premium model.

Screenshot [SCREENSHOT: aigen-generating.png — progress indicator during generation]

A deeper look at the two stages

Why research comes first

Most AI writing tools fold research and composition into a single model call. It looks efficient — one request, one response — but it trades off quality. The model either leans on its training data (which is months or years stale) or it tries to browse the web in the same breath it's supposed to be writing, and both halves suffer.

Content Studio separates them on purpose. The research pass runs first, with its own prompt, tuned for factual accuracy rather than creative framing. It queries the live web, gathers the facts most relevant to your topic, and returns a structured set of bullets with context. Those facts go into the composition pass as reference material.

The result is a draft that uses today's facts instead of yesterday's training data. You see the effect most clearly when you ask for posts on fast-moving topics — pricing, regulations, product launches, market trends — where stale information is obviously wrong.

Why the brand profile matters so much

Your brand profile is the single biggest lever on generation quality. Every field flows into the composition instruction:

• Company name and description. Used for framing — the post is about your world, not generic industry concepts.
• Industries, products, customers. Keeps the draft on-domain. A hospitality brand's "weekend trips" post won't drift into unrelated consumer advice.
• Tone and style. Determines whether the draft sounds conversational, professional, or academic; detailed, concise, or engaging.
• Call-to-action text and URL. Produces a closing block that points readers where you want them to go.

Filling in the brand profile properly is the single most impactful thing you can do for draft quality. A thin profile produces drafts that feel generic. A thorough profile produces drafts that feel like your team wrote them.

Overrides: when to use which

You don't always want the default. Three overrides let you stretch a generation for a specific run:

• Tone override. Temporary voice change for this post. Example: your default is "professional", but for a holiday-season wrap-up you want "conversational". Override once; go back to default for the next post.
• Style override. Temporary density change. Example: default "detailed" for long-form, but override to "concise" for a quick news-style write-up.
• Model override. Temporary upgrade. Example: your default is a fast model for bulk drafting, but for a flagship piece you want a premium model that can handle more nuance.

Overrides don't change your brand profile — they only affect the single run they're attached to.

Features in depth

Recreate from source

Reproduce an existing source page in your own voice. Useful for rewriting competitor listings, refreshing stale pages, or converting a long-form source into a brand-aligned blog.

Required: a source page and a brand profile.

Optional overrides: tone, style, model.

What the generation takes into account:

• Your company profile — name, description, industries, products, customers.
• Tone — conversational, professional, academic.
• Style — detailed, concise, engaging.
• The clean source page.
• Your existing slugs (so the new blog doesn't collide).
• Live research bullets on the topic.

Tips:

• If the source page is short (under 500 words), the research pass becomes the main input. Make sure your brand profile is filled in well — it carries the voice when the source is thin.
• Tone override is one-shot. For permanent changes, edit your brand profile.

Suggest new

Start from a source and get fresh blog ideas rather than a reshaped version. Returns a short list of suggestions, each with a title, a two-sentence description, target keywords, and a search-volume hint.

Use when: you have a single content-rich page and want to fan out into multiple angles before committing.

Tips:

• The search-volume hint is an estimate, not a real keyword-tool volume. Treat it as directional.
• Low-hint suggestions often have the most "white space" to capture. High-hint ones are competitive and benefit from premium execution.

SEO-targeted suggestions

Constrain suggestions to a specific keyword set. Given target keywords, Content Studio proposes angles most likely to rank.

Tips:

• Provide three to five tightly related keywords. More than ten causes focus to drift.
• Generic keywords produce generic suggestions. Pair "villa rental" with a modifier like "villa rental goa weekend trip" for sharper results.

Generate from a suggestion

Once you've picked a suggestion, generating turns it into a full Draft. The suggestion's title becomes the target; the AI can still tweak it during generation.

After generation, the suggestion is marked as selected and bound to the resulting blog, so you can't accidentally regenerate the same suggestion.

Per-run model override

Three options:

• Default model from your brand profile. Set once; applies to every generation.
• Per-run override. Pick a different model for this one generation — useful for flagship content or for A/B testing.
• Fast default. A cheaper model at scale for bulk drafting, with a premium model reserved for flagship pieces.

Per-run tone and style overrides

Pass a tone or style value on the generation form to override your brand profile for that run only. Great for one-off experiments — "make this post more casual" — without committing to a profile change.

What a draft comes back with

Every generation produces a draft that's ready to refine in the editor. Fields that are populated automatically:

• Title. Generally close to what the suggestion proposed, with the AI's polish.
• Slug. URL-safe, unique across your library.
• Body. Full long-form content in Markdown. Usually 1,200 to 2,500 words for typical blogs.
• Summary. Two-sentence gist used on list pages and in social previews.
• Meta description. Up to 160 characters, optimised for search-engine snippets.
• Keywords. 5 to 10 keywords tied to the content.
• Category. Single category label.
• Tags. Several tags for taxonomy.
• Author. Defaults to the brand profile; overridable in the editor.
• Estimated reading time. Calculated from word count.
• Cross-link suggestions. The AI's proposal for which of your other blogs would be good to link to. Not inserted automatically — use the editor's Cross-link button to merge them in.

Everything is editable in the editor after generation.

Generation strategies that work

Build a library, not a one-off

The output of generation compounds. A single draft is useful; a library of 20 drafts creates SEO authority. Plan in batches:

1. Scrape your most important pages or a set of competitor listings.
2. For each source, decide whether you're doing Recreate (reshape), Suggest new (fan out), or SEO-targeted.
3. Generate drafts for all of them in one sitting.
4. Refine in the editor over the following days.
5. Cross-link the library once several blogs are published.

Pick the right flow per source

• Recreate when the source is already close to what you want, but in someone else's voice.
• Suggest new when the source is content-rich and you want multiple distinct blogs from it.
• SEO-targeted when you have a specific keyword cluster in mind and the source is just a starting point.

Use overrides deliberately

Don't override on every run — it undermines your brand consistency. Use overrides when:

• You're writing for a different audience (a partner blog, a seasonal campaign).
• You want to A/B two styles against each other for the same source.
• You're producing flagship content where a premium model is worth the cost.

Keep the default profile doing the heavy lifting for bulk work.

Roles & permissions

Connections — cross-module workflows

flowchart LR
    A[Source scraped] --> B{Choose path}
    B -->|Recreate| C[Recreate from source]
    B -->|Fan out| D[Suggest new blogs]
    B -->|SEO-targeted| E[SEO-targeted suggestions]
    D --> F[Pick a suggestion]
    E --> F
    F --> G[Generate]
    C --> H[Draft]
    G --> H
    H --> I[Editor refines]
    I --> J[Push to marketing site]

What generation reads

• A clean source page from Sources & Scraping.
• Your brand profile from settings.
• Your existing blogs' slugs (to avoid duplicates).

What generation produces

• A Draft blog that opens in the Editor.
• For Suggest flows: a set of Suggestions saved against the source. Only one Suggestion becomes a blog.

Limits you'll see

Errors & FAQ

Common situations

FAQ

• Why do some blogs have a similar opening line? When the brand profile is thin and the source is short, the AI leans on templating. Fill in the profile more fully, vary tone/style per run, or try a different model.
• Can I re-run generation on the same source? Yes. A new Draft is created; the prior one is untouched. Clean up old drafts manually.
• Where do I see which model was used? In the editor header near the status badge.
• Why does the search-volume hint change between runs? It's an AI estimate, not a live keyword-tool number. Treat it as directional.
• Can I generate without a source? Not today. All generation paths require a source. To start from scratch, scrape even a one-paragraph source page first.
• Does generation populate cross-links automatically? No. Generation returns suggested cross-links as data; the editor's Cross-link button is what actually inserts them into the body.

Changelog

• Apr 2026 — Per-run tone, style, and model overrides.
• Mar 2026 — Live-web research pass before generation.
• Feb 2026 — Structured output — every draft has consistent title, slug, summary, meta description, keywords, and category.
• Jan 2026 — Three suggestion flows: Recreate, Suggest new, SEO-targeted.

Related modules & next steps

• Editor & Composition — where every generated Draft lands for refinement.
• Sources & Scraping — how the clean source pages are produced.
• Publishing to CRM — how drafts become live pages.
• Media Generation — turn a draft into an avatar video.