Flexa SEO – AEO | AI SEO & Answer Engine Optimization

Changelog

0.6.0

• Security: a password-protected post could have its contents read by anyone, with no password. The first 160 characters of the body were published as the meta description and repeated in the Open Graph and Twitter tags, /llms.txt carried a 150-character extract next to the URL, the XML and HTML sitemaps advertised the post, and a post using Details blocks had its questions and answers published in full as FAQ structured data, with no length limit at all. All of those now treat a protected post as having nothing quotable, and the sitemaps leave it out. If you want protected posts listed in the sitemap anyway, because only the body is meant to be private, add_filter( 'flexa_seo_aeo/sitemap/include_password_protected', '__return_true' ); brings that part back.
• Fix: the site served two sitemap indexes at once, WordPress’s own at /wp-sitemap.xml and this plugin’s at /sitemap.xml, listing the same content under two sets of URLs and leaving search engines to pick. WordPress’s is switched off now whenever the plugin’s sitemap is on, so there is one index to submit. The flexa_seo_aeo/sitemap/keep_core_sitemap filter keeps both if you need them.
• Fix: a generated description could run two words together where one paragraph or heading ended and the next began, so a post opening “What is QA?” followed by “It is testing.” produced “What is QA?It is testing.” Block boundaries become a space before the tags are stripped. Words inside a sentence, around <strong> or a link, still close up the way they should.
• New: every readiness check is now labelled documented, observed or heuristic, in the block-editor sidebar and on the Dashboard, so you can tell a requirement a search or AI vendor publishes from a reasoned bet of ours.
• New: the full scoring rubric and a tested compatibility matrix are published at https://flexa-seo-aeo.vercel.app/. The rubric gives every weight and threshold and the reasoning behind them; the matrix records what actually happens when this plugin runs next to Yoast, Rank Math, All in One SEO or SEOPress, naming the versions tested and the duplicate tags each pairing produces.
• Fix: the readme described one-click migration from SEOPress, which the plugin does not do. Migration covers Yoast and Rank Math; the SEOPress claims are gone.
• Fix: every page carried two <meta name="robots"> tags, WordPress’s and this plugin’s, and when they disagreed the one a crawler honours was down to reading order. There is one tag now: the plugin writes through the wp_robots filter that WordPress provides, so its per-post noindex and nofollow boxes merge with whatever core and other plugins add instead of competing with it.
• Fix: with Settings > Reading set to discourage search engines, the plugin still published index, follow, which is the opposite of what that box asks for and the stronger signal of the two. Discouraging search engines now forces noindex and nofollow on everything, whatever the per-post settings say.
• Fix: the Breadcrumbs and Image SEO toggles did nothing. Both sat in the settings screen, both counted towards the Dashboard score, and turning them on changed no output at all. They work now. Breadcrumbs gives you a [flexa_breadcrumbs] shortcode, a flexa_seo_aeo_breadcrumbs() template tag and a matching action, plus BreadcrumbList structured data built from the same trail, so the markup and the structured data can never describe different paths. Nothing is placed automatically, because a theme that already prints a trail would end up with two. Image SEO fills a missing alt from the attachment’s alt text, then its title, then the post title, leaving alt="" alone because that is the correct way to mark a decorative image. Stored content is untouched either way.
• Fix: a title or description migrated from Yoast that used %%name%% came out blank on posts and pages. The author name is now available there, not only on author archives.
• Fix: the plugin only warned about a conflicting SEO plugin during the Setup Assistant, which runs once, and only knew about Yoast and Rank Math. It now notices SEOPress and All in One SEO too, and shows a dismissible warning whenever one of them is switched on, since two SEO plugins means two canonical tags and two sets of social tags with search engines left to choose. Dismissing it is remembered per combination, so it comes back if you activate a different one later.
• Fix: deleting the plugin left two options behind, the Dashboard score history and the anonymous id used by the deactivation survey. Both are removed now. Your per-post SEO fields are still kept on purpose, and the FAQ above says so and explains how to clear them yourself.
• New: /llms.txt, the XML sitemaps and the Markdown version of a post now send Cache-Control, ETag and Last-Modified, and answer 304 when a crawler already has the current copy. Less bandwidth and less work for your server on the URLs bots re-fetch most.
• New: /sitemap_index.xml and the post-sitemap.xml style URLs that Yoast and Rank Math publish now redirect to the matching Flexa sitemap instead of returning 404, so a site that migrates keeps the sitemap URL it already submitted to Search Console.
• Fix: the sitemap post type and taxonomy settings accepted names that no longer exist, so the stored value could drift into something the settings screen could not display or clear. Both lists are now checked against what is actually registered.
• Fix: the REST endpoint behind the editor sidebar answered differently for a post id that exists and one that does not, which let someone without permission work out which posts are real. Both cases now give the same answer.
• Fix: a WooCommerce product or category page carried two BreadcrumbList blocks, the plugin’s and WooCommerce’s, describing the same page by different paths. WooCommerce’s is now dropped wherever the plugin emits its own, and kept untouched where it does not, so switching Breadcrumbs or Schema off hands the structured data straight back to WooCommerce.
• Fix: the BreadcrumbList on a 404 or a search results page was keyed to the home page URL, which made every one of those views claim to be the same thing. Neither view has a canonical of its own, so neither emits the block any more.
• Fix: post type archives, the WooCommerce shop page and the date archives shipped no rel="canonical" at all. WordPress prints one only on single posts and pages, and the plugin had nothing to put there, so those pages were left to the search engine’s own guess. Each now points at itself, which also gives the BreadcrumbList block on those pages a stable id.
• Fix: IndexNow handed search engines URLs that nothing else on the site advertises. A post you marked noindex, a password-protected post and every post on a site set to discourage search engines were all pushed the moment they were saved, and because the ping fired on each save, one editing session could submit the same URL a dozen times. The checks the sitemap already applies now apply here too, and a URL just submitted is not submitted again for fifteen minutes. The flexa_seo_aeo/indexnow/resubmit_after filter changes that window.
• Fix: page 2 and beyond of a blog, category, tag, author or shop listing told search engines it was page 1, and carried page 1’s title. Each page now points at itself and says which page it is, so the later pages can be indexed instead of being treated as copies. A static front page needs the opposite care, because it answers for /page/2/ and /page/99/ alike with the same body: it is numbered only as far as the content really splits, and WordPress decides where that is, so a crawler is never handed an endless run of URLs over one page. A canonical you set by hand on a post still wins.
• Fix: the sitemap index dated each file from the newest published post of that type, including posts the file leaves out. Marking your most recent page noindex left the index advertising a date no URL inside it carried. The date now comes from the posts that are actually listed.
• Fix: a WooCommerce product page carried two Product blocks in its structured data, this plugin’s and WooCommerce’s, both claiming the same identifier and disagreeing about the contents. WooCommerce’s now stands down wherever the plugin emits its own, and is left alone where it does not, so turning Schema or WooCommerce AEO off hands the markup straight back.
• Fix: the Dashboard scored pages you had marked noindex and pages behind a password, so work you had deliberately hidden from search engines dragged the site score down. The scan now reads the same list the sitemap does.
• Fix: archive titles carried the heading prefix WordPress adds for the page itself, so a search result read “Archives: Shop” or “Month: May 2026”. The prefix is gone from the title tag; the heading on the page is unchanged.
• Fix: a title template whose tagline or any other token resolved to nothing left the separator behind, so a site with no tagline could ship its home page as “Northbank Press – – Page 2” and an archive as “Shop – “. An empty token now takes the separator in front of it with it, in all nine separator styles. A title that genuinely contains the separator character is untouched: “Re–run the build” on a site set to a hyphen still reads the way you wrote it.
• New: the Organization or Person block in your structured data has its own Logo field, under Settings > Organization. It used to borrow the default share image, which is a wide social card rather than the square logo search engines ask for. If you leave the field empty it falls back to your theme’s logo, then to the share image as before.
• Fix: a shortcode belonging to a plugin that is switched off, or not loaded on that request, was published as the page’s own words. [contact-form-7 id="12"] came out as the meta description, repeated in the Open Graph and Twitter tags, in the BlogPosting description and in the /llms.txt line. Anything shaped like a shortcode is now removed from a generated description, whether or not the plugin that owns it is running, and a page left with nothing to say ships no description tag rather than an empty one.
• Fix: with the knowledge block set to Person, the structured data gave it a logo, a property schema.org defines for an organization and not for a person. A Person now carries the same picture as image instead. The Organization block is unchanged.
• Fix: the Dashboard and the editor sidebar failed a page on its meta description even where the plugin itself declines to write one: a page built from shortcodes or a page builder has no text to quote, and a protected post is never summarised on purpose. Those pages are a warning now, and the hint says which of the two it is and what the page will actually ship.
• Performance: settings are read dozens of times while a page is built, and each read rebuilt and re-checked the whole stored option. They are now resolved once per request. A legacy sitemap URL that has to be redirected no longer builds the entire sitemap it was only testing for emptiness.

0.5.5

• New: the Dashboard and Settings > Tools now point you at Flexa Cache, our free page cache. Flexa SEO builds your tags, schema and sitemap entries while WordPress puts a page together, and a page cache is what stops that happening again for the next visitor. The button opens the WordPress.org listing in a new tab, so nothing is installed behind your back; if you already have Flexa Cache sitting there switched off, it activates it instead. The Dashboard card can be closed for good; the Tools row stays. Neither appears if Flexa Cache is already running, or if you have white-label mode on.
• Fix: uninstalling the plugin left the Setup Assistant’s progress behind in the options table. It is removed now, along with the new dismissal flag.

0.5.4

• Fix: on a site with a very large number of posts, building an XML sitemap could use a lot of memory. The step that skips posts you marked noindex collected every one of those IDs first, with no limit, then repeated the list on each sitemap page. It now stops collecting past a thousand and asks the database to filter instead, so the cost stays flat however big the site gets.
• Fix: /llms.txt still used the slower post meta lookup that XML sitemaps dropped in 0.5.3, so it was both heavier than it needed to be and able to disagree with the sitemap about which posts are indexable. Both now share the same exclusion.

0.5.3

• Fix: the Dashboard content scan was slow, and on sites with long posts or a page builder it could time out before the report appeared. Scoring a post ran the whole WordPress content filter chain, once per post, so page builders, galleries, embeds and related-post widgets all fired up to fifty times over for a report that only reads text. The scan now reads the stored content directly. Scores are unchanged.
• Fix: every XML sitemap query joined the post meta table twice to skip the posts you marked noindex, which made sitemaps slow to build on large sites. That exclusion is now a single indexed lookup, reused across the whole sitemap request.
• Fix: a page holding Details blocks rebuilt its FAQ structured data on every visit, re-reading and re-rendering those blocks each time. The result is now stored and reused until you edit the post, and a page with no Details block skips the work entirely.
• Fix: the “Before you go” survey shown when you deactivate the plugin now offers a Get help link, so you can reach support instead of only leaving feedback.

0.5.2

• Fix: author, search, 404, date and post type archives carried no meta description at all, which also cost them their Open Graph and Twitter descriptions. Each of those views now gets one, built from what WordPress already holds: the author’s bio, the description a post type was registered with, or a short sentence naming the archive and the site. Category, tag and custom taxonomy pages still use the term description, and now fall back the same way when a term has none.
• Fix: a blog index with nothing written on it (a separate Posts page, or a front page set to your latest posts) also came out with no description. It falls back to the tagline, then the AEO site summary, then a plain line naming the site.
• New: a flexa_seo_aeo/metas/description filter, so you can replace any of this wording with your own.

0.5.1

• Fix: the front page could end up with no meta description at all. When the homepage is a static page, its own SEO fields were skipped, so unless you had filled the global Home description (or your site tagline), the description tag was dropped entirely, and Open Graph and Twitter lost their descriptions with it. The front page now falls back to the page’s own SEO description, then its excerpt, then the tagline. The same page’s SEO title, canonical, Open Graph image (including the featured image) and noindex/nofollow boxes are now honored too, instead of being ignored.
• Fix: a separate Posts page never produced a meta description under any configuration. It now uses its own SEO description, then its excerpt, then the site tagline, and honors its SEO title, canonical, Open Graph image and robots settings.
• Fix: pages carried two <link rel="canonical"> tags, one from this plugin and one from WordPress itself. WordPress’s copy is now removed whenever the plugin outputs its own, so a custom canonical set on a post is no longer contradicted by a second tag pointing at the default URL.

0.5.0

• New: a Links settings section for cleaning up URLs and controlling link behavior on the front end. Strip Category Base removes /category/ from category archive URLs (with a 301 from the old URL); Redirect Attachments sends media attachment pages to their parent post, with an optional fallback URL for attachments that have no parent. It also adds rel="nofollow" to external links, rel="nofollow" to links pointing at external image files, and target="_blank" to open external links in a new tab. The link attributes are applied when content is displayed, so your stored content is never changed.

0.4.1

• Fix: the deactivation feedback survey no longer opens more than one “Before you go” dialog. With several Flexa plugins active at once, each bundled its own copy of the survey script and every copy added a handler to the Deactivate link, so the dialog stacked and needed one click to dismiss each copy. Each Deactivate link now opens a single dialog.

0.4.0

• New: an optional deactivation feedback survey. If you deactivate the plugin from the Plugins screen, a short survey asks why, so we can fix what is not working. It is anonymous (a random per-site id, no email or domain), runs only in the admin, and never blocks or delays deactivation. If you reactivate later, that win-back is recorded too (anonymously). Disable it all with add_filter( 'flexa_seo_aeo/deactivation_survey/enabled', '__return_false' );. See External services for exactly what is sent.

0.3.0

• New: Setup Assistant, a guided onboarding wizard. It detects your site, recommends the SEO and AEO settings that fit, and applies them in one click. The flow covers search essentials, AI readiness, a quick content-readiness scan of your recent pages, optional migration from Yoast or Rank Math, and a final readiness report that seeds your Dashboard. Nothing is ever overwritten: recommendations are compared against the defaults, you see every setting that will change before applying, and any value you already customized is kept.
• New: re-enter the Setup Assistant any time from the Dashboard card, from Settings › Tools, or from the command palette. A re-run only proposes settings that are still at their defaults, so it is safe to run repeatedly.
• Security: running a full site scan now requires the “edit others’ posts” capability (Editors and Administrators), so lower-privileged roles can no longer trigger the site-wide scan through the REST API. Viewing the cached Dashboard report is unchanged.

0.2.0

• New: SEO/AEO health Dashboard: overall SEO, AEO and Technical scores, an issues overview, SEO & AEO health breakdowns, prioritised recommended actions, and a “Pages needing attention” list. Each listed page expands to show exactly which readiness checks fail (with a plain-English fix), links straight to the editor, and can be re-scanned on its own after an edit, with no full site scan needed.
• New: one-click “Enable site-wide” fix: readiness checks that fail only because a global toggle is off (structured data, llms.txt, Markdown alternate) get a Fix button on the dashboard and in the block-editor sidebar; it flips the setting and re-scores your pages. Shown only to users who can manage settings. No AI, no external calls.
• New: score trend, a lightweight history of your SEO/AEO scores, captured each time you run a scan (no cron, no external analytics).
• Fix: migration source detection could run an extremely slow database query on large sites (a multi-key meta lookup), which under a small PHP-FPM pool could tie up all workers. Rewritten as a single indexed query so opening the plugin (and running a migration) stays fast.
• Improved: the plugin now opens on the Dashboard, with a Dashboard / Settings switch in the header.

0.1.0

• Initial release: Answer-Engine Readiness Score (per-post 0–100 grade + actionable checklist in the block editor), title/meta templating, Open Graph & Twitter, canonical & robots, XML/HTML sitemaps, robots.txt, IndexNow, llms.txt, Agent Readiness Markdown export, JSON-LD schema (Article/WebPage/WebSite/Organization/FAQ), one-click migration from Yoast and Rank Math, optional WooCommerce Product schema, and a React admin app.