=== AnswerRank Auto AEO ===
Contributors: answerrank
Tags: ai, seo, content, aeo, writing
Requires at least: 6.4
Tested up to: 7.0
Requires PHP: 8.1
Stable tag: 1.1.0
License: GPLv2 or later
License URI: https://www.gnu.org/licenses/gpl-2.0.html

Research-driven AI articles optimised for AI answer engines and traditional search. Bring your own OpenRouter and fal.ai keys.

== Description ==

By [answerrank.so](https://answerrank.so). Support: support@answerrank.so.

AnswerRank Auto AEO writes articles the way an editorial team does: research before planning, planning before drafting, and editing after both.

Nothing generates until you save a context profile, so articles sound like your brand instead of like generic AI content.

**The method**

1. **Research** — what the question actually asks, the terms and definitions that belong in a complete answer, figures with real sources behind them, and what the pages already ranking cover well or leave open.
2. **Plan** — an outline built to be answered rather than padded: question headings, the point made early under each one, and room for the internal and external links the piece needs. Experience and authority come from the author identity you save, never invented.
3. **Draft** — the article, written to your saved house style so it reads like your publication.
4. **Edit and proofread** — a readability edit for sentence length, rhythm, repetition and filler, then an SEO and AEO proofread covering headings, links, keyword use, and whether the piece is ready to be quoted in an answer.

**Editorial standards, applied to every article**

* Sentences kept short and readable, with long ones split rather than nested.
* No repeated openers, restated ideas, or paragraphs that circle the same point.
* No hype, no unsupported superlatives, no filler transitions.
* A maintained list of tired phrases kept out, extendable with your own.
* Keywords used at a natural density instead of pushed.
* Claims tied to real sources, with no invented statistics or credentials.

These are ordinary editorial standards, applied consistently rather than left to chance. Readers and answer engines reward the same things: a clear answer, stated early, that can be trusted and quoted.

**Output is native WordPress**

Articles are built from core blocks: paragraphs, headings, lists, tables, quotes, images, details, and groups. The highlight feature uses the editor's own highlight format, so you can change or remove it like any other formatting. There is no custom block and no lock-in. Schema for Article, FAQPage and HowTo is generated and stored on the post. If Yoast, Rank Math or SEOPress is active, the meta title and description are written into their fields instead of being duplicated.

**Everything runs in the background**

Generation never runs inside your browser request. A job is handed to a detached
process, WP-Cron picks it up if that never lands, and a watchdog restarts anything
that stalls. Close the tab, reload the page, or walk away: the article keeps
writing, and the progress screen reattaches to whatever is still running.

**Works with your existing plugins**

Meta titles and descriptions are written into Yoast SEO, Rank Math, All in One
SEO, SEOPress, The SEO Framework or Slim SEO, whichever is active, so you edit
them where you already expect to. When one of those is running, the plugin stops
outputting its own Article markup so you never ship two competing graphs. On a
classic-editor site the article is stored as plain HTML instead of block markup.
Polylang and WPML get a language assigned, and WP Rocket, LiteSpeed, W3 Total
Cache, SG Optimizer and WP-Optimize get purged for the post.

**Images**

A featured image plus a configurable number of images inside the article, all in
the visual style you chose during setup. Each one lands in the media library with
a title and descriptive alt text, placed after the section it illustrates.

**Autopilot**

Set a cadence, a publishing window and a weekly cap, then choose whether autopilot saves drafts or publishes directly.

Autopilot keeps its own topic queue full. When the queue runs low it works out what to write next from your brand context and everything already published on the site, skipping anything you have covered before. Connect AnswerRank and it draws from measured gaps first. You can still add topics yourself, and the queue page shows what is waiting, where each topic came from, and lets you reorder or remove anything you do not want written.

**Expand your AEO with AnswerRank**

The plugin writes. AnswerRank decides what is worth writing. Connect your account and the questions where AI assistants are not yet recommending you flow into the topic queue, so every article aims at a measured gap instead of a guess. The plugin writes without an account too.

== External services ==

This plugin sends data to third-party APIs that you connect yourself. Nothing is sent until you add a key.

* **OpenRouter** (https://openrouter.ai) — receives your topic, your context profile, and the article text at each pipeline stage in order to generate research, outlines, drafts and edits. Terms: https://openrouter.ai/terms. Privacy: https://openrouter.ai/privacy.
* **fal.ai** (https://fal.ai) — receives an image prompt derived from the finished article when image generation is enabled. Terms: https://fal.ai/terms. Privacy: https://fal.ai/privacy.
* **AnswerRank** (https://answerrank.so) — optional. Receives your brand ID to fetch content opportunities and score data. Only used when you connect a key.

During the site scan, the titles, URLs, excerpts and headings of your published posts and pages are sent to OpenRouter so it can summarise what your site covers. No post content beyond the excerpt is sent.

== Installation ==

1. Upload the plugin folder to `/wp-content/plugins/`, or install the zip through Plugins, Add New, Upload.
2. Activate it. The setup wizard opens automatically.
3. Add an OpenRouter key, optionally a fal.ai key, describe your brand and author, and let it scan your site.

For scheduled publishing on a low-traffic site, add a real cron entry rather than relying on WP-Cron:

`*/15 * * * * cd /path/to/wordpress && wp cron event run --due-now`

== Frequently asked questions ==

= Do I need an AnswerRank account? =

No. It is optional and everything works without it.

= Where do API keys go? =

They are encrypted with libsodium before being stored, using a key derived from your site's `AUTH_KEY` salt. They are never returned to the browser after saving, and only a masked version is ever displayed.

= How much does an article cost? =

You pay OpenRouter and fal.ai directly, and the plugin shows a live estimate on
the writing screen plus the real spend per article once it is done. Rates come
from OpenRouter's own catalogue, so the figures follow whatever the provider
currently charges.

A 1500-word article runs about 38,000 input and 15,000 output tokens on the
default Balanced setting, plus about 12,000 more if live web search is on.

The model you assign to each role matters far more than anything else. Using one
premium model for all three roles costs roughly ten times more than using a fast
model for research and editing and keeping the premium one for writing only. The
estimate on the writing screen updates as you change models, so try a few.

Live web search is on by default, because current facts are the one thing a model
cannot supply for itself. The search returns whole pages rather than snippets, and
every page is billed as input at roughly 20,000 tokens, so the cost depends almost
entirely on which model does the research. On a fast model that is under two cents
per article; on a premium model it is ten times that. Untick it per article for
evergreen topics using the switch on the writing screen.

Cited sources are verified with a HEAD request before publishing either way, so a
URL that does not exist is unlinked rather than published.

Avoid models that run their own web search, such as Perplexity's Sonar family, for
the research role. They ignore the page limit entirely and can put a single call
over 150,000 input tokens.

An image adds $0.003 to $0.06 depending on the fal model.

= It takes several minutes per article. Can it be faster? =

Settings, AI has a speed control with Fast, Balanced and Maximum depth options,
which trade research depth against time. The editorial standards apply either
way. The biggest lever is the model itself, since a heavy reasoning model can
take two minutes for a single call. Turning off live web search removes another
20 to 40 seconds.

= Can I edit the prompts? =

Yes. Settings, Advanced has an override box for every stage.

= Do I lose my settings when I upgrade? =

No. Upgrading replaces the plugin files only. Your API keys, context profile,
author identity, writing rules, topic queue and article history all live in the
database and are untouched.

New settings introduced by an upgrade appear with their defaults, and anything
you had already changed is kept. The schema is repaired on the first page load
after an upgrade rather than on activation, so replacing the files is enough and
you never need to deactivate first. Articles that were mid-generation resume on
the next watchdog pass.

Uninstalling is the only thing that can remove data, and only if you turned on
"Delete all plugin data" in Settings, Advanced beforehand.

= Will the articles be flagged as AI content? =

The plugin removes the mechanical tells that detectors and readers pick up on, and it enforces those rules with code rather than hoping. It is still your responsibility to review, fact-check and take editorial ownership of anything you publish.

== Screenshots ==

1. The setup wizard: connect OpenRouter, choose models, describe the brand.
2. Writing an article, with live progress and per-stage cost.
3. A finished article in the block editor: quick answer box, takeaways, FAQ.
4. The topic queue that feeds autopilot.
5. Settings: writing rules, SEO and AEO controls, autopilot schedule.

== Changelog ==

= 1.1.0 =
* Autopilot now finds its own topics when the queue runs low, instead of stopping and emailing you.
* New "Find topics now" button on the Topic queue page.
* The queue shows where each topic came from.
* New settings: turn automatic topic discovery off, and choose how many topics to keep waiting.

= 1.0.0 =
* First public release.
