What you are selling
One sentence you can say without thinking, then the evidence behind it, read from the Austen repository rather than from the marketing site. Everything marked "built and working today" was verified in code on 4 September 2026 and rechecked on 30 September 2026.
Austen turns a brief into a finished, on-brand article that is optimised for search and for AI answer engines, and a professional editor at Proofed edits every one before you see it.
- You approve a brief. Nothing is written until you have.
- Austen researches, drafts, scores against SEO and GEO rubrics, rewrites itself against its own scorers, enforces the word budget and strips AI writing tells.
- A Proofed editor edits the draft and returns tracked changes you accept or decline line by line.
- The finished article publishes to WordPress, or to anything else through a signed webhook, and spawns images, social posts, newsletters and translations.
What it does, in the buyer's words
| They say | Austen does | Built and working today |
|---|---|---|
| "AI writing all sounds the same." | Brand voice is inferred from a URL, from uploaded samples or from what they already publish, then applied to every draft. Per-project style guides come from Stylus. A tell gate removes AI writing tells before delivery, and readers can flag phrasing in the editor, which bans it for that project immediately. | Yes: tone-of-voice inference, Stylus style guides, config/prose_tells.php, reader flags with an admin queue for global enforcement. |
| "I do not trust an AI draft to go out unread." | A professional editor at Proofed edits every article and returns it as tracked changes on the original. You see what changed, not just the result. | Yes: fulfilled through Proofed's live API, billed per word at $55 per 1,000 words with a $15 minimum.S2 |
| "Nobody finds us in ChatGPT any more." | Every article is scored against a GEO rubric as well as an SEO one, and improved against both scorers before anyone sees it. The article shows both scores, with a note when either is below the 90 target, rather than passing it off as finished. | Yes: ArticleQualityGate, target 90, up to two improvement passes per article. Most articles still finish below 90 on at least one score, so never promise it.S6 |
| "We do not know what to write about." | Discover competitors, crawl and analyse their sites, produce a content gap report, and convert a gap straight into a brief. Keyword research is included. | Yes: app/Services/Crawl, competitor analysis, gap-to-brief. |
| "Our old posts are dead." | One "Improve an existing article" entry: give it a URL or paste the text, then choose how much may change. Light polish and Thorough optimization return tracked changes on the original, accepted or discarded one at a time. Refresh and Rewrite analyse the piece, pre-fill a brief the buyer approves, and write a new version through the full pipeline. Works on external content. | Yes: /projects/{project}/improve, over Optimize import and Content Refresh. A refresh raises a performance gap only when Search Console data backs it. Pushing a refreshed piece back to the same WordPress post is part of the admin-only CMS sync beta. |
| "We already said all of this on a webinar." | Upload audio or video, transcribe with speaker diarization, edit the transcript, and write the article from it so the piece carries the team's own arguments. | Yes: source assets, transcription and summaries. |
| "We publish in five markets." | Translations into 32 languages with right-to-left support for Arabic, Hebrew and Persian. | Yes: app/Enums/SupportedLanguage.php. |
| "Getting it onto the site is the slow bit." | Scheduled publishing, WordPress destinations via an application password (a WordPress draft by default), signed webhook destinations, and a two-way WordPress sync through the Austen Connect plugin in admin beta. | Yes for publishing and webhooks, with WordPress publishing working since the fix of 23 September 2026; the plugin and two-way sync are admin-only beta. |
| "We run content for eleven clients." | One project per client, each with its own brand voice, style guide and competitors. An agency account can also join a client's own workspace without using one of that workspace's seats. | Yes. Project limits are per plan (Pro has five), and each workspace is billed on its own plan; agency accounts are flagged by an admin.S1 |
| "Can my assistant just do it?" | An MCP server at austenapp.com/mcp with five tools, so Austen works as a connector in Claude, plus a REST API documented at /api-docs, whose tokens can be limited to one project. The connector drafts briefs and writes articles; ordering the human edit and publishing stay in the app. | Yes: key-free OAuth registration, no token to paste.S6 |
The pipeline, because it is the answer to most objections
Every article written from a brief, whether it comes from the app, the API, the MCP server or a Refresh of an existing article, goes through one code path: ArticleGenerationPipeline::generate(). Both the real brief flow and the admin model bench call it, so the two cannot drift.S6
| Step | What happens | Why a buyer cares |
|---|---|---|
| 1. Research | Live web search grounds the article in real sources before a word is drafted. | It is not writing from memory. Say this when someone raises hallucination. |
| 2. Draft | One structured generation on the power model against the brief, the plan, the brand voice and the research notes. | The brief is the contract. Anything wrong here was wrong in the brief they approved. |
| 3. Section self-check | A deterministic, free check that every heading the plan required is present. On a miss it generates only the skipped sections and splices them in. | The article covers what the brief promised, not what the model felt like writing. |
| 4. Quality gate | Scores SEO and GEO, then improves the draft against the scorers' own fixes, up to two passes. A draft that plateaus is kept and flagged. | Two published scores, not a vibe. And an honest flag rather than a silent pass. |
| 5. Length gate | The brief's stated word limit is the contract; the plan's target is the fallback. A draft more than 15% over is trimmed. | Nobody pays for 2,400 words when they asked for 1,200. |
| 6. Tell gate and clean | AI writing tells are removed. Em dashes and en dashes are stripped by ArticleProse::clean() as a hard brand rule enforced by tests. | The single most common giveaway in AI copy does not survive to the page. |
| 7. Human edit | Proofed editor, tracked changes back on the original, with the editor's comments shown in the same review. | This is the step no competitor in the software category has. |
What it is not, yet
Say these plainly when asked. Overclaiming costs the second meeting, and in this category the buyer has usually already been oversold once.
- There is no customer you can name, no testimonial and no case study. The development database has zero users, zero articles and zero subscriptions, and this guide had no production access.S7 Do not imply otherwise anywhere, ever.
- Credits are measured, not enforced. CREDITS_ENFORCE is false, so a heavy account is visible in /admin/costs (and in credits:usage, which arrives with the pilot branch) but nothing is deducted or blocked. Selling volume before you read that report is how a retainer loses money quietly.S7
- The pilot SKU is on a branch. gtm/pilot-offer adds the /pilot page, the guest Stripe checkout, the pilot:sync-price command, the provisioning webhook and the buyer's status page. Until it is merged, austenapp.com/pilot does not exist. Once merged, until STRIPE_PILOT_PRICE_ID is set, the page renders and the button honestly says the pilot is not open for card payment yet. The branch was cut on 4 September and main is now 71 commits ahead, so merging it means a rebase and a retest.S3
- A score of 90 is a target, not a promise. The product dropped its "90+ SEO and GEO pass" line on 30 September 2026 because it was not true: in the 60 days before, 11 of 35 articles reached 90 for SEO and 3 of 35 for GEO. Show the two scores; never quote a threshold.S6
- The two-way CMS sync is an admin-only beta behind a feature flag. Do not sell it. One-way publishing to WordPress works, but only since a fix on 23 September 2026; before it, no article had reached a WordPress site, so test a publish on the buyer's own site before you promise a date. The pull-optimise-push loop is not a promise yet.
- No SSO. The pricing page lists SSO on Enterprise, a contact-us tier with no self-serve price, but it is only a label: the code has no SAML or OIDC integration, only sign-in with a Google or Proofed account. Tell a buyer SSO is not available yet; if they need it, it is a build with a date, not a feature.S1
- Human editing is priced per word and is a real cost. It is not bundled into Starter or Pro. On the pilot and the retainer you bundle and hide it; on a subscription the buyer sees it as an add-on.
- Austen writes articles. It is not a design tool, a video editor, a social scheduler beyond its own posts, a link-building service or an analytics product. It connects Google Search Console; it does not replace it.
The stack, so you can answer "where does our content go"
Read from the repository, not from the marketing site.S34
| Layer | What | Say this |
|---|---|---|
| Application | Laravel 11, PHP 8.3, Alpine.js, Tailwind, Tiptap editor, database queue | "A standard, well-supported web stack." |
| Login | Passwordless: a magic link that expires after 15 minutes, or sign-in with a Google or Proofed account. No password login. Turnstile on the magic-link form. | "There is no password to leak, because there is no password." |
| Generation | Anthropic and OpenAI models, routed per job purpose; article generation runs on the power model, scoring on the standard modelS5 | "What we write from goes to the model: your brief, your brand voice, and any source material, existing page or Search Console figures you ask us to work from." |
| Human editing | Proofed, over its API, per project server; each order asks for Proofed's Proofreading service by default, with the brief and an instruction to humanise the copy. It is a human edit, not a fact-check | "The same editorial firm, with the same editors, that handles our managed clients." |
| Style guides | Stylus, provisioned per project | "Your style guide lives as structured rules, not as a PDF nobody opens." |
| Billing | Laravel Cashier and Stripe; secrets encrypted at rest; webhook signature validation | "Card details never touch us. Stripe holds them." |
| API and assistants | Sanctum API tokens, either account-wide or limited to one project (read or write), and Passport (OAuth2) tokens, which are account-wide; idempotency keys on writes; MCP over the same auth, with no project limit enforced | "An API token can be limited to one project, so it cannot read another client's work." Do not say this of the Claude connector. |
| Outbound integrations | Webhooks signed HMAC-SHA256; every user-supplied URL and every page Austen reads or crawls passes an SSRF guard, and a redirect is either re-checked or refused | "Signed, so your CMS can prove the request came from us." |
| Mailgun for transactional mail | Factual; only say it if asked. |
Prompt injection. User content is sanitised and length-limited before it reaches a model, with clear delimiters between instructions and content. This is worth volunteering to any buyer who is technical, because most of the category cannot answer the question at all.
Who it is for
Agencies who resell content
Five to 150 people, billing clients for content programmes, sourcing writers freelance. Buyer: founder, head of content, delivery director. Trigger: an open writer vacancy, which is a public admission that delivery capacity is short. Austen already supports the workspace shape they need.
In-house content teams of one
B2B SaaS, education, professional services. Buyer: head of marketing, or the founder in a 20-person company. Trigger: an open content req. The email writes itself, because the pilot costs less than a month of the salary being advertised.
Regulated and claims-heavy brands
Supplements, skincare, financial and legal explainers. Buyer: head of content, with a medical or compliance reviewer behind them. Trigger: an unedited AI draft is unusable to them, which makes the human edit the entire purchase.
One line against each alternative
| They are considering | Say |
|---|---|
| ChatGPT or Claude directly | "You already have the draft. What you do not have is a brief that was approved, a scored article, a professional editor's tracked changes and a record of what was published. That is what the price buys." |
| Byword or Frase at a few dollars an articleS11S12 | "They sell drafts and they are very good value for drafts. Nobody there edits your article. If you can publish their output unread, buy it." |
| Verblio's AI plus human hybrid at about $90 an articleS21 | "Closest thing to us on paper, and honestly priced. The difference is that our editor works on your style guide and returns tracked changes on your file, and our brief is approved before anything is written." |
| A freelance writer at $0.10 to $0.14 a wordS22 | "A good freelancer beats us on a single flagship piece. We beat a freelance bench on ten pieces in 30 days, because you brief once and chase nobody." |
| A content agency retainer averaging $3,209 a monthS26 | "Our done-for-you retainer is $2,000 to $2,500 for eight finished, human-edited articles a month, and you keep the workspace if you ever want to bring it in house." |
| Hiring a content marketer | "Hire them. Then use the pilot to fill the four months before they are productive, and give them a working engine on day one." |
| Jasper or Writer at enterprise ACVsS14 | "Different purchase. They sell a platform to a marketing organisation. We sell finished articles to a team that needs them next month." |