> For the complete documentation index, see [llms.txt](https://thecontentforge.gitbook.io/thecontentforge-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://thecontentforge.gitbook.io/thecontentforge-docs/feature-glossary/glossary/content-optimization.md). # Content Optimization *Last updated: 2026-08-07.* The learning loop. It studies your own results, what the platforms are rewarding right now, and what your competitors are doing, then turns all of it into specific changes to what you post next. ## Brand Voice How your brand sounds, written down once and applied to every draft. * **Brand Voice** - A saved, editable definition of how your brand sounds: tone rules, patterns to do more of, things to avoid, signature moves, real example lines, hook patterns, closing patterns, and the themes that consistently perform. Every draft the product writes is held to the active version. * **Vocabulary and emoji style** - Alongside the rules, a list of words to prefer and words to avoid, plus a free-text note on how emoji should be used, covering frequency, placement, and which ones belong to the brand. * **Generate from post history** - Writes a first draft of your voice from your imported posts that actually performed, so it reflects what worked rather than how you think you sound. * **Generate from project knowledge** - Builds a voice from your onboarding positioning, your project profile, and the documents you flagged as safe for content, which is the way in for a project with little or no post history. If there is not enough material yet, the product lists exactly which fields to fill in, highest impact first, and where to add them. This source is offered to Web3 project accounts; every other account generates from post history. * **Blend** - Runs both sources and merges them into a single voice, drawing on each where it has the most to offer. If your project material is too thin, the voice is built from post history alone and the product tells you what was missing. * **Edit voice guidelines** - Every line can be edited, added, or removed by hand, and each is marked strict, meaning the writer must follow it, or suggested, meaning it is optional. Lines are labelled by origin, so you can see at a glance which came from you, which the AI proposed, and which came from your project material. * **How posts should start** - A list of required openers such as BREAKING or Hot take. Entries you add yourself are enforced as a hard rule during generation, while AI suggestions stay advisory. * **Length preferences** - Set a maximum word count per platform for X, Instagram, and Facebook, or leave them blank and let the writer choose based on the post. * **Platform voice guidelines** - Write a different tone and structure per platform, for example punchy and hook-first on X but longer and more editorial on Facebook. Leave one blank to use the general voice there. * **Voice versions** - Each generation is saved as a new numbered version with a date. You choose which one is active, and old versions can be deleted by an owner or admin once another version is active. * **Version allowance** - A bar on the page shows how many saved versions you have against your plan's limit. Every version counts, including old ones you no longer use, so deleting superseded versions frees up room to generate again. * **Social guidelines** - Longer freeform guideline documents sit alongside the voice and can be read and edited in place, for cases where a rulebook does not fit into short lines. ## Patterns What your own published posts show about what works, recomputed from your real numbers rather than asserted. * **Pattern Engine** - Recomputes your performance patterns on demand over a window you choose, from seven days up to all time, and produces one card per platform you have data for. Everything on the card is calculated from your own posts, so the numbers are auditable rather than opinion. * **What's Working Now** - A summary strip across the top: for each platform, the best-performing hook, the best-performing length, the best closing, the sample size behind it, and whether engagement is up or down over the last week. * **What to do next** - Three concrete actions per platform, covering the hook to use, the length to target, and the closing to add, each with the measured lift behind it and a Draft button that opens the writing tool pre-loaded with that instruction. * **Benchmarks** - Median, 75th percentile, and 90th percentile for each metric you have data on, plus engagement rate, so you know what a good post looks like for you specifically. * **Engagement fingerprint** - Shows how your best posts earn engagement: likes, comments, shares, saves, or follows, and labels the account accordingly. Useful because a comment-heavy audience and a save-heavy audience need different content. * **Hook Templates** - Detects the opening structures you actually use, such as breaking-news openers, numbered lists, question hooks, or contrarian takes, and ranks them by lift against your own baseline. Expand any one to see your best real examples and draft a new post using that structure. * **Length Performance** - Performance by character-count band, switchable between engagement score, likes, comments, shares, saves, and follows, with sample sizes so you can judge confidence. * **Topics** - Groups your posts into the topics you actually write about, and for each topic shows the hook, length, and closing that worked best. This is how you tailor format to subject rather than using one template everywhere. * **CTAs** - Every closing style detected in your posts, scored against posts with no call to action, with average comments, shares, and follows for each. It shows which asks are actually earning replies rather than which ones feel natural. * **Features** - Compares posts with and without tickers and mentions, and posts with zero, one to two, or three or more hashtags, so you can settle those arguments with your own data. * **Drift** - What changed recently: which openers you have started leaning on, which ones you have dropped compared with your recent baseline, and whether your engagement rate has moved enough to be worth reacting to. * **Top Posts** - Your highest-scoring posts in the window, expandable to full text with the metrics, detected hook, closing, and length, and a copy button. * **AI Pattern Analysis** - A deeper pass over the computed data that runs automatically after a recompute and can also be re-run on its own. It returns what is working with the evidence behind it, what is not working with a suggested fix, an overlooked opportunity, an optimal post formula covering hook, length, closing, topic mix and features, named patterns with why they work and a reusable template, prioritised recommendations, and notes on brand voice consistency. * **Use Today** - Ready-made post templates produced by that analysis, each with the context it suits and real examples, plus a copy button and a Draft button that carries the template into the writing tool. * **Algorithm-fit feedback** - Sits under the AI analysis and sorts your patterns into keep, retire, and try, with the reason for each and a label showing whether the judgement came from the platform brief, one of your own logged algorithm notes, or simply from what was observed in your posts. Advisory only, and it never overrides the core analysis. * **Follower Growth Playbook** - Per-platform tactics ranked high, medium, and low priority, each with the detail, the reasoning, and the expected impact, plus a second view listing the specific data points from your own account that the tactics lean on. * **How to read this page** - A collapsible explainer with four cards covering what the hook, length, topic, and AI analysis panels measure and how to read lift and sample size. ## Experiments Deliberate tests, with published results collected for you and the lesson recorded at the end. * **Experiment Tracker** - A running list of the tests you have going, with counts for active, waiting for results, completed, and lessons captured. Each row shows the hypothesis, platform, success metric, and status. * **Experiment templates** - Eight starting points that prefill a hypothesis, a metric, and two contrasting variant strategies: hook test, call-to-action test, posting time test, short versus long format, educational versus urgent tone, question versus statement ending, competitor-inspired angle, and visual style test. They are offered while the tracker is still empty; after that you start from the New experiment button. * **Five-step builder** - Walks you from hypothesis, to platform and content type, to writing the variants, to choosing a primary metric with optional secondary metrics and a measurement window, to saving as a draft or attaching scheduled posts immediately. * **Up to four variants** - Each variant carries a name, a short statement of what makes it different, and the actual content. Variant A is the baseline the rest are compared and generated against. * **Generate an alternate variant** - With your Variant A written, one click drafts a genuinely different take on it for the B, C, or D slot. Nothing is saved until you review it. * **Attach posts** - Tie existing drafts or upcoming scheduled posts to a specific variant, either during setup or later from the experiment itself. This is what connects a test to real published performance. * **Automatic results** - Once attached posts publish, their real metrics flow into the experiment on their own, updating regularly and every time you open it, with a Refresh metrics button when you do not want to wait. Any cell can be overridden by hand when a number has to come from somewhere the product cannot reach. * **Variant comparison** - A side-by-side table of every metric across every variant, with the primary metric highlighted, manual entries marked as such, and a clear note while the test is still waiting for published data. * **Winner detection** - The leading variant on your primary metric is marked automatically on the variant cards, so the result is obvious without reading the table. * **Learning capture** - When a test finishes you record the winning variant, a one-line takeaway, what specifically drove the result, and when to use it again, and Suggest writes a first draft of all of it for you to edit. A tick box pushes the recommendation into your Brand Voice signature moves on save, which needs an owner or admin to actually apply. * **Test a draft from the writing tool** - A draft can be sent from the writing tool straight into the experiment builder, arriving as Variant A with the platform already set, so testing something you just wrote takes one step. * **Status control** - Activate a draft, mark it completed, or archive it. Only active tests count against your plan's experiment allowance. ## Algorithm Watch What each platform is rewarding right now, kept as a rulebook with the evidence attached and fed into generation as optional context. * **Algorithm Watch** - Tracks how X, Instagram, and Facebook are actually ranking content right now and feeds that into every generation as optional context. The product is explicit that these signals are advisory: your brand voice, intent, tone, and goal always take priority. * **Signals** - The default view. It shows today's brief per platform, grouped into trends, do more of, avoid, and format tips, with a confidence level, a confirmed-or-experimental label, and a link to the source behind each claim. Alongside it sit the count of monitored sources and active rules, per-platform scan health, and any rule changes from the last 24 hours. * **Signal sources** - The blogs, accounts, newsletters and forums behind the rulebook, each tagged by what kind of source it is, and reachable from the Data Ingestion area as well as from here. The set that ships with the product is kept current for you and cannot be edited. Sources you add yourself are flagged for you to review instead, the page reminds you when one is overdue, and you can mark one reviewed, switch it off, or remove it. * **Active Rules** - The full current rulebook per platform. Each rule shows the claim in plain English, its category, confidence, whether it is confirmed or experimental, its status, how many sources back it, and when it last changed. Filter by platform, by category, or by state: active only, experimental, changed in the last seven days, weakened, or deprecated, and expand any rule for notes, when it was first seen, when it was last confirmed, and how much evidence sits behind it. * **Org Updates** - Log an algorithm change your team observed yourself, with a summary, the action it implies, a suggested test, a confidence level, and a source link. Recent entries are taken into account when the product writes for this account, and older ones are kept but marked archived. * **Current Events** - The news the AI writes from, browsable directly. Filter by time window from twelve hours to a week, by category such as crypto, political, geopolitical, finance, and regulation, and by source, or search the text. You can pull the latest feeds on demand, copy the whole set to the clipboard, and switch to a trackers panel that keeps the last five items per watched theme. * **Run a scan now** - Owners and admins can trigger a source refresh on demand and watch it progress, rather than waiting for the next automatic update. ## Competitor Intelligence What the accounts you watch are doing, turned into angles, tests and gaps you can act on. * **Competitor Intelligence** - A dashboard that turns competitor movement into content decisions. The header shows whether the current analysis is fresh or stale, how many sources you track, how many of their posts are held in the rolling ninety-day window, and when the last scan ran. * **Run Intelligence Scan** - Fetches the newest posts from every active source and re-runs the analysis in one action. If nothing could be fetched, it names the specific sources that failed and why, so a broken feed is easy to spot. * **Manage Sources** - Add competitors as X profiles or feeds, with a name and optional notes, then enable or disable them individually. The drawer shows how many active sources your plan allows and warns clearly when you are near, at, or over the cap, and over the cap only the first allowed sources feed the scan. Deleted sources sit in a recently deleted list for three days so a mistake can be undone. * **Market overview** - A row of counters for formats, hooks, narratives, tests, things to avoid, and posts, each jumping straight to its panel. * **Narrative Momentum** - A chart of the recurring themes across your competitors over seven, thirty, or ninety days, so you can see which story is gaining and which is burning out. Where engagement data is not available for a source, the score is an estimate and is labelled as such. * **Top Formats** - Which post formats competitors are leaning on right now, as a share of their output, each with a trend direction, why it matters, and a recommended action. * **Winning Hooks** - The opening patterns landing across the accounts you watch, with a real example, why it works, how to use it, and a signal-strength rating. Each hook can be sent onward as a draft, a counter, or an experiment. * **What To Test** - Concrete experiment ideas derived from the scan, each with a priority, the reasoning, a suggested post structure, and the platform it applies to. * **What To Avoid** - Patterns underperforming across the market, each with a risk level and a safer alternative, so you do not copy something that is already fading. * **Market Gaps** - Topics competitors touch but leave room on. Each gap shows the evidence from their posts, what is actually missing, the angle you should take, and a starter draft idea. * **Competitor Posts** - A feed of the recent posts feeding the analysis, tagged with format, hook type, and theme, with links out to the originals and a note of how many more are retained. * **Per-competitor drilldown** - Open any single competitor for a written summary, their top hooks, formats, and narratives, their strengths, where they are weak or repetitive, and the counter angles open to you. Below that sits every post from that source in the window, tagged the same way. You can refresh just that source or switch it off from here. * **Create Experiment** - Sends any insight, hook, gap, or whole competitor into the experiment builder pre-filled with a title, the reasoning, and a starting variant, so testing an idea takes one click instead of a retype. * **Draft a counter** - Sends an insight into the writing tool as a one-off instruction to write an original post that responds to that angle in your own voice, with an explicit guard against imitating the competitor or reusing their wording. * **Send to Content Forge** - Saves an insight to your stored source-context list so the angle is there to pull up later. It is a bookmark rather than a live instruction: use Draft a counter when you want the angle to shape the next generation. * **Export Brief** - A one-click export of the current intelligence as a shareable brief. (beta / in progress) * **Competitor Analysis (legacy)** - The original single-shot competitor page: add sources on the left, run one analysis on the right, and read a list of suggestions you can push into Brand Voice or into an experiment. It still works and stays up to date, but it has no navigation entry and is kept only as a fallback. Competitor Intelligence supersedes it and covers everything this page does with far more depth. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://thecontentforge.gitbook.io/thecontentforge-docs/feature-glossary/glossary/content-optimization.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.