| 开发者 | oxfordmetadata |
|---|---|
| 更新时间 | 2026年8月23日 19:15 |
| PHP版本: | 7.4 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPLv3 or later |
| 版权网址: | 版权信息 |
[aieo_semantic_related] shortcode; either can be narrowed to one kind of content, so a listing recommends listings rather than products.ProductGroup + hasVariant, per-variant colour, gtin, brand, shipping and return policy.
* AI titles and descriptions written against the live SERP. Drafts await approval, or run on Autopilot.
* Opportunity grid — traffic × conversion: hidden gem, intent mismatch, defend, prune.
* Rank tracking (optional DataForSEO, under your spend cap).
* Redirects and 404s, slug monitor, static sitemap, robots.txt, alt-text filling, EU Omnibus history, hreflang, Local SEO.
Selling more
wp-content/plugins/.No, and nobody does. Everything is off by default. Rooster One installs beside your current SEO plugin, pixel manager and consent tool and changes nothing at all until you switch a piece on. Turning one thing on does not commit you to the next. The right way in is to pick the single problem that is costing you most right now and switch on only that. Most shops start with recommendations or the AI chat, see what it does for a fortnight, and only then look further. There is no configuration you must complete before anything works.
Because it is one idea, not twelve features. Rooster One measures your whole business on a schedule — every product, page, customer, order and session, scored into flat tables — and every feature reads those same tables. That is why the recommendations, the SEO engine, the audiences, the ad ROI and the AI assistant agree with each other instead of each maintaining a private version of the truth. Twelve plugins cannot do that: they would each need their own copy of your catalogue, their own idea of who your best customers are, and their own cron job to build it. It is also why the plugin gets faster to use as you switch things on rather than slower. The expensive thinking already happened, on a schedule, off your customer's request. The second feature you enable reads figures the first one already paid for.
Ctrl+K from anywhere in wp-admin reaches every addressable place in the plugin, pages, tabs and individual panels, searchable by what a thing does rather than by what it is called. All Features lists every one with pinnable favourites, and every page carries a written description of what it is for. If you cannot remember where something lives, you are not expected to.
Yes. Pro customers get direct support, and there is a support group one click from the top bar in wp-admin for everyone. Beyond support we offer paid set-up and configuration, and further customisation where a shop needs something the plugin does not do out of the box. Worth saying plainly: the plugin is built for a real shop we run ourselves, so what a working shop needs tends to already be in it, and when it is not, we are the people who can add it. If you would rather hand the configuration to somebody who has done it before, that is a service we sell rather than a favour we occasionally do.
Yes, and this is the case we design for rather than tolerate. The plugin is GPLv3, so there is nothing to license and nothing obfuscated. Your team gets 218 filters and 48 actions with the aieo_ prefix covering the parts most likely to need bending, a REST surface, and nine WP-CLI commands for anything that belongs in a deployment script rather than in a browser.
Behaviour is switchable in code as well as in the admin, so a decision your team wants to own can live in your own plugin instead of in our settings table. Where an integration is genuinely missing we would rather add the hook than have you patch the plugin, since a patched copy stops receiving updates. Send us the case.
We are also happy to work alongside an in-house team rather than instead of one: they keep the domain knowledge and the deployment pipeline, we cover the parts of the plugin nobody outside it should have to reverse-engineer.
No. Rooster One works on content sites and catalogues too, and adapts what it measures: a content site is scored on visits, engagement and form submissions rather than sales. A badge in the header tells you which of the three it has decided you are, so you are never guessing why a number looks the way it does.
No. It is a complete product: the AI assistant, the full SEO engine, the whole pricing and merchandising stack, the storefront features, measurement and consent, product feeds and courier fulfilment are all in the free version. Pro adds operational capability, not the basics. The withdrawal button is free as well, because it is a legal obligation for every EU distance seller and charging for it would be a poor thing to do.
The operational half: marketplace order import, attribution and ROI, audiences and automated campaigns, web push, customer intelligence, Insights Chat, and events and competitions. Those screens are already in this download and a subscription switches them on. Pro also includes two companion plugins you install separately — Rooster One Runtime, which serves the storefront without loading the admin half and is the biggest single speed difference between free and Pro, and Rooster One Pro, which carries supplier invoice scanning, the accounting and ERP connectors, the catalogue importer, the product exchanger, the duplicate-product merge engine and semantic internal linking.
Because two different things are called Pro, and only one of them arrives when you connect. The trial and the subscription grant an entitlement, which unlocks the Pro screens whose code is inside this plugin — AI chat, managed AI credits, attribution, audiences and the rest. Six features live in the separate Rooster One Pro plugin: invoice scanning, the accounting and ERP connectors, the catalogue importer, the product exchanger, the merge engine and semantic internal linking. An entitlement cannot switch on code that is not on the server, so those screens tell you to install the plugin, which your subscription includes. Install and activate it and they come alive with no further setup.
No, and that is deliberate. Pro gates the screens where you curate and configure, never the code that draws your shop. The semantic related-content section, the recommendation rails and the policies you have already set keep rendering from what was already worked out, on the free plugin, with no Pro plugin loaded at all. What you lose is the ability to change them: the Duplicate products screen, Semantic Internal Linking and the feed-source generator stop opening, and their controls stop answering. We would rather a lapsed licence cost you the settings screen than blank a live section of your website in front of your customers.
No. The plugin is GPLv3 and a great deal of it — the chat, recommendations, the SEO engine and every storefront and pricing tool — runs without any subscription. Only features that actually call a model consume Rooster AI credits, and the ones that do say so before they run rather than after.
The design goal is the opposite. Expensive work is pre-computed on a schedule into flat tables, and the storefront reads single cached rows, so enabling a feature adds a feature rather than a per-request query. On Pro the runtime plugin serves the front end without loading the admin half at all.
That is what the architecture is for. Heavy analysis runs on a schedule into flat pre-computed tables; the storefront reads single cached rows regardless of how many products sit behind them. Background jobs re-arm themselves after updates, stay dormant until a feature is actually used, and will collapse a runaway third-party task rather than queue politely behind it.
Questions about your catalogue, grounded in it. It recommends only real, in-stock products and links to them exactly, because it answers from your data rather than from a general model's memory. You control a knowledge base for the things a catalogue cannot tell it — opening hours, delivery times, returns policy — and identity guardrails stop it describing your business as something it is not.
It says so, and then tells you. Chat analytics show what was asked, how answers were rated, which questions found nothing, and which conversations ended in an order. The list of questions that found nothing is usually the most valuable screen in the plugin: it is your customers writing your content backlog for you.
It reads the same prices the storefront does, including role and wholesale pricing, so a logged-in trade customer and a retail visitor are told different and correct numbers. It is not working from a snapshot taken when the model was trained.
Chat runs on Anthropic Claude and semantic search on Voyage AI embeddings. Nothing else leaves your site for those features. Insights Chat, which answers questions about your own shop inside wp-admin, pseudonymises customers and exports no data at all. Every external service the plugin can talk to is listed under "What does Rooster One send to an outside service, and when?" below.
It can. Semantic search indexes products, posts, pages and terms as embeddings, so "something for dry hair in winter" finds the right products without those words appearing in any title. It runs alongside the built-in search rather than ripping it out, so you can compare before you commit.
From what your customers actually did, not from tags you maintain by hand. Up-sells, cross-sells and same-need alternatives are derived from real order history and refreshed on a schedule, so a new product earns its place as soon as it starts selling and a discontinued one leaves without you remembering to remove it.
Yes. You order the rails by drag and drop, they work on both block and classic themes, and custom slots let you place one anywhere a theme accepts a block. Each rail can be turned off independently, which matters more than it sounds: the commonest cause of a slow product page is four rails nobody asked for.
Yes, and this is the part most plugins leave out. Lift measurement reports the incremental orders and revenue the engine adds, rather than counting every order that happened to contain a recommended product. The second number is always bigger and always flattering, which is presumably why it is the one usually reported.
Yes. The pricing engine handles role and wholesale prices, brand-level discounts, loyalty tiers, points and free gifts, and it stacks them in a defined order rather than letting the last filter to run win. Prices are resolved ahead of time into the same pre-computed tables everything else reads, so a complicated rule set does not become a slow product page.
Yes, with colour, image and button styles, and they render on archives and product pages from a cache rather than by loading every variation. Alongside them sit labels, trust badges, payment logos, gallery handling, a layout builder, custom tabs, and one master button style so the shop stops looking like four plugins bolted together.
Carts are saved server-side, and the restore is cookieless so it survives CDN caching and a customer moving device. Recovery messages go out through automated campaigns by email or web push, using WordPress's own mail path or your SMTP plugin. You do not need a third-party marketing service, and quiet hours are honoured so nobody is pushed a cart at three in the morning.
Win-back, birthday, back-in-stock, post-purchase follow-ups and reactivation, all triggered from the same audience data. Audiences themselves — top customers, lapsed, lookalike, abandoned-cart, reactivation, competition entrants — are recomputed nightly, so a campaign aimed at "lapsed" means lapsed today rather than lapsed whenever the segment was last built by hand.
From behaviour the plugin already records, recomputed nightly rather than assembled by hand. The ready-made ones are top customers, lapsed, lookalike, abandoned-cart, reactivation and competition entrants, so "lapsed" means lapsed as of last night rather than lapsed whenever somebody last exported a spreadsheet.
Yes. The audience builder takes rules over what customers bought, browsed, abandoned or responded to, combined the way you would say them out loud: ever purchased this brand, not in the last ninety days, opted in. An audience is a live definition, not a frozen list, so it repopulates itself as people move in and out of it.
Send it a campaign, target an offer at it, exclude it from another campaign, or push it as a segment for advertising. The same definitions feed the ROI reporting, so you can ask what a segment actually returned rather than only how many people are in it.
No. Push is served from your own site with its own opt-in gating, and the same quiet-hours rules apply. It is part of the campaign engine rather than a separate product with its own dashboard.
No. The SEO engine self-gates behind AIOSEO, Yoast and Rank Math and stays dormant until you explicitly switch it on. There is a read-only importer and a head-parity check, so you can compare the output before cutting over, and switch back if you do not like what you see.
It knows what you sell. Commerce schema is assembled from the same tables the rest of the plugin uses: ProductGroup with hasVariant, per-variant colour, gtin, brand, shipping and return policy. Titles and descriptions are written against the live search results page and wait for your approval, or run on Autopilot if you would rather they did not.
That is exactly what the head-parity check exists to answer. It compares what your current plugin emits against what Rooster One would emit, page by page, before anything changes. The importer is read-only, so your existing plugin's data is never altered, and the engine stays gated until you decide.
The opportunity grid plots traffic against conversion and sorts pages into hidden gems, intent mismatches, pages to defend and pages to prune. That is a different question from "what is my score out of a hundred", and a more useful one. Underneath it sit redirects and 404 monitoring, a slug monitor, a static sitemap, robots.txt, alt-text filling, hreflang, Local SEO, and EU Omnibus price history.
Optionally, through DataForSEO, under a spend cap you set. It is off unless you turn it on and give it your own credentials, because rank tracking is the one part of SEO with a per-query cost and we would rather you decided the size of that bill.
Yes, with full block and classic parity. That is deliberate and tested rather than assumed: the newer block checkout behaves differently enough that features working in one can silently do nothing in the other. Several bugs we have reported upstream to WooCommerce were exactly that shape.
Yes. The plugin holds sixteen carrier integrations and talks to each with your credentials; it is not a reseller and takes no cut of your shipping. Vouchers, labels, cash on delivery, tracking, bulk printing and daily closings are handled in wp-admin, and a manual carrier covers anyone without an API.
Delivery economics computes the real break-even from your actual carrier costs rather than from a round number that felt about right. It also reports profit and loss per carrier and per zone, which usually reveals that one zone has been quietly subsidised by the rest for a long time.
Yes: invoices, packing slips, proformas and credit notes, singly or merged across a date range for a whole day's dispatch. They are generated from the order rather than from a template you keep in sync by hand.
Both directions. Ready-made feed templates cover the major marketplaces and shopping engines, with a custom builder for anything not on the list, and marketplace orders come back as native WooCommerce orders rather than into a separate silo. The catalogue importer runs the same road in reverse when you want to pull a marketplace listing set in.
It complements them and makes them survive ad blockers and consent choices. Conversions go server-side to Meta CAPI, GA4, Google Ads Enhanced Conversions and the TikTok Events API, deduplicated against the browser tags you already have, sent after the order is saved so a failed send can never cost you an order, with a daily safety net that re-sends anything that slipped.
A first-party consent banner with Google Consent Mode v2, a proof-of-consent log so you can show what a visitor agreed to and when, and a cookie scanner that reports what your site actually sets rather than what you believe it sets. Measurement respects the answer rather than working around it.
The ROI Wizard reports orders, revenue, cost, net, ROAS and ROI by engine, offer, channel, campaign or creative. Ad spend can be typed in, uploaded as CSV, or synced nightly from Google Ads, Meta and TikTok. Shops that do not sell can set conversion goals instead and be measured on those.
Yes. The composer turns what the shop already knows — your product sets, brands, categories and demand signals — into a complete draft Google Ads campaign, written down in full before anything touches Google's API. You read the campaign, the ad groups and the actual copy as plain text and decide.
No, and this is the part we would want you to check rather than take on trust. A plan has three states. Draft exists only on your site. Created means it has been pushed to Google Ads and is sitting there paused, real objects spending nothing. Activated means a human enabled it in Google. Rooster One records that third step. It never performs it. There is no code path in the plugin that turns a campaign on, which is deliberate: composing an advert and committing your money are different decisions and only one of them should be automatable.
Ask the customers who already told you. Content gaps are distilled from the storefront chat agent's own conversations: the questions that returned no products, and the answers customers rated badly, weighted by how often they come up. That turns "what do people keep asking that we answer poorly" into a ranked list of things to write, sourced from demand rather than from a keyword tool's guess. It is the loop we are most pleased with in the whole plugin: the chat widget's failures become the SEO team's brief.
In two steps, kept separate on purpose. First it ranks which of your products and posts are worth promoting, using the same demand scores the SEO engine uses, so you are not promoting whatever happened to be published last. Then, for one object you choose, it drafts platform-native copy for Facebook, Instagram, X, LinkedIn or TikTok. It only ever drafts. Nothing is published, and no social account tokens are touched by the drafting step. Ranking candidates is free; writing the copy is the part that consumes AI credits.
Somewhere to ask your shop a question in plain language inside wp-admin, instead of building a report. Customers are pseudonymised and no data is exported. It is the one AI feature aimed at you rather than at your visitors.
Stock recovery finds sales you lost to out-of-stocks, missing-hot-sellers finds products you should be carrying, and an out-of-stock predictor warns you before the shelf empties. There is branch stock for multi-location retail, barcode scanning, and an optional ERP bridge.
Nothing is removed. Tables are never dropped on update, activation or deactivation — only when you explicitly uninstall the plugin. Deactivating to test something is safe.
Yes. Admin and storefront are both fully translatable, a complete translation ships with the plugin, and source strings are English and extractable with Poedit.
It gives you the mechanism Directive (EU) 2023/2673 requires: a permanent, prominent withdrawal button that works for guests, a separate confirmation step, partial withdrawal on multi-item orders, and an email acknowledgement recording the content, date and time of the declaration. That is the part software can do. Every EU member state transposes the Directive into national law, so check the date that applies in your own market. The rest is yours: your terms of sale, who pays return postage (Directive 2011/83/EU article 14 leaves that partly to your terms), and actually processing the declarations and refunds you receive. The plugin will not pretend those are done because a button exists. The feature ships disabled for that reason — switch it on when your policy is ready.
No, and treating them as the same is the commonest way a compliant-looking implementation fails. Cancelling an unshipped order is a commercial courtesy you grant on your own terms and may refuse once the parcel has left. Withdrawal is a statutory right exercisable for 14 calendar days from receipt, which you cannot refuse and whose window you do not set. Rooster One keeps them as two separate flows.
Rooster One runs on your own server. It contacts an outside service only in the cases listed here, and every one of them is either something you switched on or something you pressed Send on. rooster.systems — our own service. Privacy policy: https://rooster.systems/privacy/
For WooCommerce stores, product recommendations are produced by our graph-database service at https://eshop-optimizer.com. It is privacy-centric: you can opt out of sharing product and category names and prices with limited impact on quality, and withhold customer IDs for anonymity. Oxford Metadata Ltd maintains privacy provisions approved by both Google and Facebook — see https://eshop-optimizer.com/compliance. Hashed PII is never stored or logged raw. Customers are pseudonymised in Insights Chat. Consent is recorded with an IP/UA-hashed proof-of-consent audit log.
101.55.3* takes only codes starting with it, *-EL only those ending with it, and a plain EL- any code containing it. A dot stays a dot, so real codes behave the way you type them.