AI content automation

AI that writes, illustrates and publishes to your sites

Because Zinn Digital® hosts the site, AI acts on it directly. Build a recipe — a trigger and a set of steps — and it runs on a schedule as a durable workflow: write the post, generate the image, publish through the site's WordPress API, update the record, notify you. Bring your own model keys, set a spend cap, and decide whether anything goes live without your approval.

Start free See pricing

100,000+PBN sites hosted across every niche
5prebuilt recipe types in the library
58languages content translates into
$19–39per month, AI Content Assistant

The difference between drafting and publishing

Most AI writing tools hand you text and stop at the clipboard. Ours doesn't stop there, because the site is already on our infrastructure — the same engine that provisions it, caches it and backs it up also runs the automation that fills it.

An AI Pack recipe reaches all the way through: it drafts the copy, generates the imagery, pushes the post through the site's WordPress API, writes the result back to your records and sends you a notification. There is no export step, no plugin to authorise, no separate integration to keep alive between your writing tool and your host.

That end-to-end reach is what turns AI from an assistant you supervise into a schedule that runs. You define what should be published, how often and against what guardrails — and then it happens, whether or not you are at your desk.

A recipe is a trigger and a set of steps

Recipes are yours to compose. Each one pairs a trigger with an ordered chain of actions, and each run executes as a Temporal workflow — durable, retryable and idempotent, so a failed step resumes rather than duplicating work or leaving a half-published post behind.

The example we ship as the canonical pattern is deliberately mundane and deliberately complete: every N days, write a post about X, generate an image, publish it via the site's WordPress API, update the record, notify the owner. Everything else in the library is a variation on that shape.

Trigger on a schedule — every N days, weekly, monthly — so a site keeps publishing without anyone remembering to.
Chain steps in the order you want: write about a topic, generate an accompanying image, publish, update the database, notify.
Runs are logged action by action, so you can see exactly what the AI did on which site and when.
Long-running work never blocks a page or a dashboard — automations execute as background workflows, not in a request.
Because runs are idempotent, a retry after a transient provider error resumes the recipe instead of publishing the same post twice.

A prebuilt library to start from

You do not have to compose a recipe from nothing. Five recipe types ship ready to run — point one at a site, set the cadence, adjust the prompt and guardrails, and it is live.

Auto-blog

Keep a site publishing on a fixed cadence. Give it a topic, a schedule and a tone, and it writes and posts through the site's WordPress API without a manual step in the middle.

Image generation

Produce featured images and in-post artwork with your own image-model key, generated as part of the same run that writes the post rather than sourced separately afterwards.

Alt text

Generate descriptive alt text for images across a site — the accessibility and on-page detail that gets skipped at volume, handled automatically.

SEO rewrite

Rewrite existing pages and posts against SEO intent rather than starting from scratch, so pages that already have history get improved instead of replaced.

Translation

Translate published content into the locales you serve, wired into the platform's translation lifecycle so a source change marks the translations stale and re-translates them.

Your keys, your models, your bill

The AI Pack runs on bring-your-own keys. You hold the relationship with the model provider and you pay your own inference cost — we run the automation, not a marked-up resale of someone else's tokens.

BYO provider keys

Add your Anthropic, Google, OpenAI or image-model keys to the credential store, held in Vault. They are never written into code, config files or logs, and they stay yours.

A model registry, not a hardcoded list

Providers, models and their capabilities live in an admin-editable, auto-synced registry. Model IDs are data — when a provider ships a new model you can select it, rather than waiting on us to redeploy.

Everything through the AI gateway

All model and embedding calls ride the AI gateway, which gives cost and token analytics, rate limiting and provider fallbacks — so a single provider having a bad day doesn't stop your schedule.

Caching applied by default

The gateway caches automatically: stable prompt prefixes aren't re-billed on every call, and deterministic outputs like translations and embeddings are cached by content hash so unchanged content is never re-processed.

Guardrails, because it publishes on its own

Automation that can spend money and put content live in public needs limits designed in from the start. Four of them apply to every recipe.

Per-customer spend caps

Set a ceiling on what your automations may spend against your provider keys. Caps bound the whole AI Pack, not one recipe at a time.

Optional human approval before publish

Turn on approval and a recipe drafts everything, then waits. Nothing reaches the live site until you have read it and said yes. Leave it off and the schedule runs unattended — the choice is per recipe.

Action logging

Every action an automation takes is logged, so you can reconstruct what was written, generated or published on any site, and when.

Rate limits

Requests are rate-limited at the gateway, which keeps a runaway loop or an over-eager schedule from burning through your provider quota.

Written once, published in every language you serve

Content created or edited on the platform is translatable across all 58 supported languages, and translation is wired into the content lifecycle rather than treated as a manual afterthought.

Create or edit a piece of content and the platform marks each locale for translation automatically — you don't queue anything by hand.
Change the source and its translations are marked stale, then re-translated, so localised copy doesn't quietly drift out of date behind the English.
Translations are cached by content hash, so unchanged text is never paid for twice.
The same lifecycle covers the surfaces that usually get forgotten — metadata, structured data text, alt text and translated URL slugs on public pages, not just body copy.
Translation status is visible per locale, so you can see what's current, what's stale and what's still pending.

Grounded in the SEO data we already hold on your sites

The AI Content Assistant sits alongside the rest of the SEO toolkit, so what it writes is informed by what we already measure about the site it's writing for.

Every plan includes free SEO Basics on each site — indexed status, Authority Score and homepage Core Web Vitals. Add the paid SEO Pro layer and you get rank tracking, backlink monitoring, competitor intel and a monthly site health audit, with the AI Content Assistant listed among those add-ons at $19–39/mo. The underlying data comes from established providers behind swappable adapter interfaces — DataForSEO as primary, Serper.dev as a cheaper failover tier and Majestic for Trust Flow and Citation Flow — so no single vendor holds your SEO stack hostage.

Because the content assistant and the SEO data live in the same dashboard as the hosting controls, a rewrite recipe can be pointed at the pages the audit actually flagged, rather than at a list you exported from somewhere else and pasted in.

Plans and add-ons are priced from the catalog and shown in your own currency at checkout, so the figure you see is the figure you pay.

One foundation, whichever way you want to drive it

AI automations are one surface onto the platform's action catalog. The same catalog is what the in-dashboard AI assistant calls, and what any MCP-capable agent reaches over our MCP server.

Run scheduled recipes from the dashboard when you want content produced on autopilot.
Ask the in-dashboard assistant to do something one-off, using your own model key through the AI Pack.
Connect Claude Code, Cursor, ChatGPT or Claude Desktop over MCP and drive the same actions conversationally from your own tools.
Every route is bounded by the permissions of the identity behind it, and every privileged action is audit-logged.
Build once, expose many ways — a capability we add to the platform becomes available across all three surfaces, with no per-tool integration treadmill.

FAQ

Whose AI account does this run on, and who pays for the tokens?

Yours, and you do. The AI Pack is bring-your-own-keys: you add your Anthropic, Google, OpenAI or image-model credentials to the Vault-backed credential store and pay your own inference cost directly to the provider. We run the automation, the scheduling and the publishing. Gateway caching is applied by default — stable prompt prefixes aren't re-billed each call and deterministic outputs like translations are cached by content hash — which is there specifically to keep that bill down.

Will it publish to my live site without me seeing the content first?

Only if you configure it that way. Human approval before publish is an option on every recipe: switch it on and the automation drafts the post and the imagery, then holds until you approve. Switch it off and the schedule runs unattended, which is the point for high-volume networks. Either way, every action the automation takes is logged, so you can review exactly what was published and when.

Can a runaway recipe empty my API credit?

There are three limits working against that, though no automation should be left entirely unwatched. Per-customer spend caps bound what your automations may spend against your keys; rate limits at the gateway stop a loop from firing requests at full speed; and gateway cost and token analytics show you what is actually being consumed, per call. Set a cap that matches your appetite and check the analytics after the first few runs of any new recipe.

Which models can I use, and what happens when a new one launches?

Providers, models and their capabilities live in an admin-editable, auto-synced model registry — model IDs are data, never hardcoded into the product. That means new models can be selected as they become available rather than waiting for a platform release. Calls route through the AI gateway, which also provides provider fallbacks if your primary model is unavailable.

How does a recipe actually reach my WordPress site?

Through the site's WordPress API, because the site is on our infrastructure. A run drafts the content, generates the image, publishes the post via that API, updates the record and notifies you. Each run executes as a Temporal workflow, so it is durable and idempotent — a step that fails on a transient provider error resumes rather than republishing the same post.

Can it write in languages other than English?

Yes. Translation is one of the prebuilt recipe types, and the wider platform translates content into all 58 supported languages as part of the content lifecycle — including metadata, structured-data text, alt text and translated URL slugs on public pages. Edit the source and its translations are marked stale and re-translated, so localised versions don't drift behind the original.

What does the AI Content Assistant cost, and can I try the platform first?

It is listed among the paid SEO add-ons at $19–39/mo, priced from the catalog and shown in your own currency. You can start on the card-free 7-day Footprint-Free trial — no payment details to begin, up to five sites — and every paid plan is backed by a 30-day money-back guarantee, free migrations and no vendor lock-in.

Put the AI where the sites already live

Start on the card-free 7-day Footprint-Free trial — no payment details, up to five sites. Bring your own model keys, set a spend cap, and let a schedule do the publishing. 30-day money-back guarantee, free migrations, no vendor lock-in.

Start free