CiteCue AI Auto-Fix -- WordPress 插件

开发者	citecue
更新时间	2026年9月17日 02:36
PHP版本:	7.4 及以上
WordPress版本:	7.0
版权:	GPLv2 or later
版权网址:	版权信息

CiteCue AI Auto-Fix is the WordPress end of CiteCue, an answer engine optimization (AEO) platform. It decides, per request, which version of a page WordPress returns — the optimized one to a recognised AI crawler, your normal page to everyone else — and it can do that only from inside WordPress, before the theme renders.

Per-request delivery to AI crawlers — when GPTBot, ClaudeBot, PerplexityBot, ChatGPT-User or any other agent in the crawler registry requests a page, the plugin returns the CiteCue-optimized version of that URL. Human visitors always see your normal site, and optimized responses are never cached for regular traffic. Any miss, timeout or outage passes straight through to the normal page.
Gap-filling page metadata — adds CiteCue's title, meta description, OpenGraph, canonical and structured-data tags to your live pages, so search engines and AI answer engines see them on the page a human sees. It fills gaps only: it reads what your theme, WordPress and your SEO plugin actually printed into <head> and adds only what none of them emitted, so there is never a second title or canonical.
Page enhancements — where you have approved one in CiteCue, adds a short facts-and-FAQ section to the end of a page you already have, built from your own answered facts and grounded FAQ entries rather than generated. It is placed only if the page does not already carry one, and it is the only thing this plugin adds that your visitors can see.
Recognises 18 AI crawlers, and learns new ones on its own — GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot, GoogleOther, Amazonbot, Bytespider, CCBot, meta-externalagent, MistralAI-User, DuckAssistBot and more. The registry refreshes daily from CiteCue, so an agent that launched last week is recognised without updating the plugin.
Never makes a visitor wait — the path that renders a page for a person reads cache only; nothing on it calls out to the network. A URL with nothing cached yet renders untouched and the fetch is queued for afterwards, so the first visitor to a cold page pays one un-enriched view rather than one slow one. Requests from AI crawlers are the deliberate exception — only a bot is waiting there — and even those are bounded by a three-second timeout, a per-minute lookup budget and a circuit breaker that stops the plugin trying at all while CiteCue is unreachable.
Content from CiteCue — a signed endpoint through which CiteCue can push new brand-building content (content briefs, FAQ packs, gap-filling pages) into WordPress as drafts for your review.
WooCommerce-aware — cart, checkout, account pages and cart-modifying links are never intercepted, while product and shop pages are served optimized. Pushed content can also create or enrich WooCommerce products (draft by default, matched by SKU with explicit consent).

This plugin requires a CiteCue account — citecue.com — and does nothing until you connect one. See "External services" below for exactly what is sent where. What the plugin actually does Everything above happens on a live request, and that is where the design decisions are:

It serves a different representation per requester, safely. Crawler matching runs against a registry that refreshes daily, so an agent launched last week is recognised without a plugin update. A logged-in user, a cart URL, a WooCommerce endpoint or a cart-modifying link is never intercepted. A circuit breaker, a per-minute lookup budget, negative caching and a stale-while-revalidate cache mean an outage at CiteCue costs a passthrough, never a broken page or a slow one.
It composes with your SEO plugin rather than replacing it. The metadata layer detects what was actually printed into <head> rather than looking for particular plugins, so it behaves correctly beside Yoast, Rank Math, a plugin nobody has heard of, or none at all. Every tag it adds carries a data-citecue attribute, so View Source tells you exactly which ones came from CiteCue.
Nothing from the API is trusted as markup. Every returned tag is parsed, matched against an allowlist of shapes and rebuilt from escaped values, with structured data re-encoded so it cannot escape its own script element.
It never makes a visitor wait on a third party. The render path reads cache only; a URL with nothing cached yet renders untouched and the fetch is queued to WP-Cron.
Content flows back in. The signed citecue/v1 endpoint is how CiteCue delivers new content into WordPress — as drafts, with replayed signatures rejected — so the loop from "this page is missing" to "this page exists" closes without anyone copying and pasting.

The service behind all of it is at citecue.com.

1.2.1

The settings screen now says where a CiteCue account comes from. Before connecting it links to citecue.com; once connected it names CiteCue as the source of what the site serves. The Plugins list gets a CiteCue.com link in the plugin's row.
This page now links to citecue.com in its own text. The directory does not turn a bare URL into a link, so the URLs in "External services" — including the terms and privacy policy — have been rendered as links rather than left as plain text.
The Author URI header points at citecue.com rather than at app.citecue.com, the login app. That is the link behind "By CiteCue" on the Plugins screen in wp-admin.
Documentation only otherwise: the description no longer leads with llms.txt, and several claims about coverage and speed were narrowed to what the code actually guarantees.

1.2.0

New: page enhancements. CiteCue composes a collapsed facts-and-FAQ section for a page you already have — built from your answered facts and grounded FAQ entries, never generated by a model — and this release places it on that page, immediately before </body>.
The block is placed exactly as CiteCue composed it. It is sanitized before it is sent and capped at 32 KB on the wire, and this plugin checks the cap again before placing it; nothing is re-escaped or run through the content filters, either of which would break the markup or run a shortcode quoted inside an answer.
Never placed twice. A page that already carries the data-citecue="page-enhancement" marker — because CiteCue's Worker fronts the site, or because the section is already in the post — is left exactly as it is.
The block rides the /seo-head response the plugin already fetches for metadata, so an enhanced page costs no extra request, and the same rules apply: cache-only on the render path, so a visitor never waits on CiteCue, and nothing at all is fetched on a site with metadata injection switched off.
Existing connections need one reconnect. CiteCue records what a plugin can do when it connects and withholds anything it never announced, so a site connected before this release is sent no blocks until it reconnects — Settings → CiteCue now says so, where before it only noticed a changed metadata setting.
"Accept pushed content" now corrects itself. If CiteCue no longer holds a signing secret for this site — because you reconnected without allowing content pushes — the switch here turns off and says so, instead of reporting a channel that cannot deliver anything. It is never switched back on remotely: that stays your decision, on this screen.
The plugin also now asks for CiteCue's site-wide identity metadata on pages that have none of their own. It is gap-filling like everything else in the head: if your theme or SEO plugin already prints structured data, CiteCue's is dropped rather than added beside it.

1.1.2

The metadata layer no longer holds an output buffer of its own open across a request. On WordPress 6.9 and later it uses core's template enhancement output buffer, so the plugin opens no buffer at all; below that it opens one in the form PHP finalizes by itself, and closes nothing. The previous shape opened a buffer on one hook and closed it on another, which left one open on any page whose wp_head did not run to the end — and a buffer left open is one the next plugin's ob_get_clean() can take by mistake.
The tags now go immediately before </head> rather than at the end of wp_head, and the check for what is already there reads the head only. Markup in the body — a <title> inside an inline SVG, a <meta> quoted in page content — no longer counts as a slot somebody else has filled, so pages carrying either get their metadata again.
The capture is arranged after every other template_redirect callback, so a request another plugin redirects or answers itself is never buffered.

1.1.1

Admin notices are confined to the Plugins screen and the CiteCue settings screen. The rejected-key warning and the duplicate-install warning used to print on every screen in the dashboard; neither asks for anything that can be done anywhere else, and the settings screen states both a second time in its status card.
The reconnect prompt can now be dismissed permanently, per user and per site. It is advice rather than an error, and an administrator who has read it and decided against it should not keep being told. On multisite the dismissal is scoped to the site it was made on, since the condition it reports on is per-site while WordPress stores user metadata network-wide.
Fixed a fatal error on a site that still has the pre-WordPress.org copy in a citecue/ folder alongside this one. That copy is 1.0.0, which predates the duplicate-install guard and loads its classes unconditionally — and WordPress always includes citecue-ai-auto-fix/ first, so 1.0.0 was always the copy that redeclared them and took the site down. This copy now stands aside for it and says which folder to delete, so the site keeps running either way.
Both duplicate-install warnings now also appear in Network Admin → Plugins. A network-activated copy can only be removed from there, and the hook they were on does not fire anywhere in Network Admin — so on multisite the warning was missing from the one screen where a super admin could act on it.
Uninstalling from a network now clears dismissal records for every site, not just the one running the uninstall. WordPress stores user metadata in a single table shared by the whole network, so anything left there is left for good.
Uninstall removes the dismissal records along with everything else.
readme: shorter summary, and the description now says plainly what this plugin does that a static llms.txt generator does not.

1.1.0

New: enriched page metadata. CiteCue's title, meta description, OpenGraph, canonical and structured-data tags are added to your live pages, so search engines and AI answer engines see them, not just AI crawlers. Uses CiteCue's /api/delivery/v2/seo-head endpoint.
Fills gaps only: anything WordPress, your theme or your SEO plugin already prints is left untouched, so the plugin never emits a second title or canonical. Detection reads the real <head> output rather than looking for particular plugins.
Never delays a page: the render path reads cache only. A URL with nothing cached yet renders untouched and the fetch is queued for afterwards. Cached blocks survive a CiteCue outage and are served while a refresh runs.
The connection now tells CiteCue whether this site injects metadata, so CiteCue stops reporting fixes as reaching human visitors when nothing puts them on the page. Sites connected before this release are asked to reconnect once.
Nothing from the response is printed as it arrived: every tag is parsed, checked against an allowlist of shapes and rebuilt from escaped values, with structured data re-encoded so it cannot escape its own script element.
Query arguments WordPress does not recognise are stripped from the URL before it is cached, queued or sent, and store pages are skipped — so order keys, reset tokens and nonces are never included, and a visitor cannot fill the scheduler with unique addresses for one page.
Changing the selected CiteCue project now clears the delivery cache, instead of serving the previous project's pages, llms.txt and metadata under the new one for up to a day.
New filters: citecue_should_inject_seo_head (skip a page), citecue_seo_head_tags (change what is printed), citecue_seo_head_query_vars and citecue_seo_head_schedule_budget.

1.0.3

No functional change. Annotates the DONOTCACHEPAGE definitions so code-quality tooling stops reporting a naming-convention violation the constant cannot avoid — page caches look for that exact name.

1.0.2

Translations no longer depend on the plugin loading them by hand; WordPress.org supplies them for the plugin slug.
The requested URL is sanitized on the way to CiteCue in a way that preserves percent-encoding, so pages with accented or non-Latin characters in the address are served and cached under the address they actually have.

1.0.1

First release in the WordPress.org plugin directory.
An install that has not been connected to CiteCue now makes no outbound requests of any kind: the daily crawler-registry refresh waits for a connection.
The plugin folder and text domain are now citecue-ai-auto-fix, matching the WordPress.org slug, so translations from translate.wordpress.org load. Settings and the connection are unaffected.
A second copy of the plugin installed in another folder stands down with an admin notice instead of taking the site down.
readme.txt now documents every request exchanged with CiteCue, and links the terms and privacy policy.

1.0.0

Initial release: AI-crawler delivery middleware (CiteCue delivery API v2), llms.txt serving, signed content-ingest endpoint, daily crawler-registry refresh, admin settings screen.
WooCommerce support: store-page exclusions for the middleware; product create/enrich through the ingest endpoint.
Hardening: single-use ingest signatures (replay rejection), per-minute delivery lookup budget, CiteCue-compatible cache-key URL normalization, cache eviction on delivery misses, and crawler-registry downgrade rejection with the bundled token floor.
One-click connect: a pairing handshake sets up the site without copying an API key in or a signing secret out, with a built-in "Verify installation" check. Connecting with an API key remains available as a fallback.

CiteCue AI Auto-Fix

标签

下载

详情介绍:

安装:

升级注意事项:

常见问题:

更新日志: