What’s new

A large quality pass gathered from live testing across 17 article types and 12 languages. Highlights since v1.0.66:

Multilingual & non-English

• Non-English articles now render the correct page-language attribute, and right-to-left languages (Arabic, Hebrew and others) display in the correct direction.
• Section labels are fully localized: the Pros/Cons boxes, the opinion disclosure line and the article-type badge now match the article's language instead of leaking English.
• Step-by-step guides no longer show a duplicated English "What you'll need" label above their materials list.
• English sources cited inside a non-English article are clearly marked "(in English)" within the link.

Structured data (schema)

• Article subtypes (NewsArticle, TechArticle, ScholarlyArticle, OpinionNewsArticle and more) are now preserved even when Yoast, Rank Math or All in One SEO manages your schema — the specific type, language and section are grafted onto their node instead of falling back to a generic Article.
• News articles carry a proper dateline, and a review with no rating is emitted as an Article (never an invalid Review).
• Scholarly articles always publish a Google Scholar author tag, so they can be indexed as scholarly work.

Citations & references

• Dead and fabricated citation links are filtered out before publishing (genuine sites that merely block bots are kept).
• Citation link text is always a readable page title — never a bare domain or a research-tool name.
• Long-form articles (pillar guides, white papers and scholarly articles) reliably get a References section built from the sources they actually cite.
• Press releases no longer attach external citations, since they are first-party announcements.

Internal linking

• Suggested internal links now use genuinely relevant anchor text (no more price figures or filler phrases), and "Insert link" works on older hand-written posts too.
• A new Internal Links page shows live scan progress across your whole content library.

Free plan

• Free articles are capped at 1,500 words; Pro unlocks the longer formats.

Pro sites update automatically from Plugins. Free updates roll out on WordPress.org.

This release polishes non-English generation based on live Japanese-site testing.

Fixed

• Image alt text on translated articles now always uses the article's own language, even when the AI briefly drafted a heading in English.
• Duplicate section headings can no longer appear when a stray English heading is translated back into the article language.
• Structural sections (Key Takeaways, FAQ, References) are now excluded from automatic images in every language, not just English.
• Headlines and page titles can no longer start with stray symbols (a leaked "│" was reaching titles and URLs).
• FAQ schema questions are now validated for shape — no more overlong or multi-line entries, and stored schema keeps its line breaks intact.

Improved

• Voice-assistant (Speakable) markup now reaches the page even when Yoast, Rank Math or All in One SEO manages your site's schema — SEOBetter attaches it to their WebPage entry instead of dropping it.

Pro sites update automatically from Plugins → Updates. Free users can grab v1.0.66 from the download page.

• Fixed: articles written in languages other than English now get relevant photos. Your keyword and the section headings are translated to English before the photo search runs, so a Japanese article about running shoes gets running shoes, not sushi.
• Fixed: badly formed headline suggestions (truncated, doubled words, broken grammar) are filtered out before they can become your post title. Checked across English, French, German, Spanish, Italian, Portuguese, Arabic and Japanese.
• Fixed: if generation is interrupted mid-article (a hosting hiccup, for example), it now stops with a Retry button instead of quietly saving a much shorter article. Articles that still land well under the requested length get a clear warning in the results panel.
• Fixed: when research finds no verified sources, the search is retried once at the end of generation. If there are still none, you see a clear warning before publishing instead of the article quietly shipping without a References section.
• Fixed: the Key Takeaways section is now always included when the article type calls for it, and never duplicated, in every language.
• Improved: local business articles now link to the exact Google Maps business card for each place instead of a search link that sometimes failed to load.
• Improved: the AI Crawler Audit now explains what it cannot see. Bots blocked at the CDN level (for example Cloudflare's AI Crawl Control) never reach your site, so the audit can show "Allowed" while the CDN is blocking. Check your CDN dashboard if a bot never shows up in your logs.
• Fixed: deleting a leftover duplicate copy of the plugin no longer wipes the active install's settings and licence.
• Improved (Pro): when a licence has expired, the update row shows a clear Renew link instead of a failed download. Security fixes remain available to everyone regardless of licence status.

• Fixed: picking a Pro content type in the generator showed "Failed to connect to API". You now get a clear message that the type is part of Pro, and your current selection is kept.

• Dead reference links are now filtered out before they reach your article. A source that returns a 404 page will not be cited.
• Cleaner citations: tracking clutter is stripped from source URLs, and social media group posts no longer appear as references.
• More relevant stock images: photos are matched against your article topic before one is chosen, so a travel article no longer gets an unrelated landmark as its hero image.

• New: a short intro video on the welcome screen and in the Settings quick-setup panel, plus a "Watch" link on the dashboard.
• Setup now clearly recommends adding a free Pexels key, with a direct link to get one. It takes about two minutes and your articles will look much better with real, relevant photos.

• Fixed: article schema included an empty author entry that failed validation. Free articles now correctly list your site's organization as the author.
• Fixed: some content types could emit an organization or profile schema node that pointed nowhere.

• Fixed: some reference links were cut short during generation and led to 404 pages. Damaged links are now repaired against the verified source list instead of shipping broken. Links that cannot be repaired are removed rather than left dead.

• Fixed: blog posts sometimes showed extra "Introduction" and "Conclusion" sections after the FAQ that repeated the opening text.
• Blog posts now get their own styling: a content-type badge and a highlighted opening paragraph, so they match the polish of the other types.
• Fixed: the score in the editor sidebar, the toolbar pill and the article metabox could disagree. All of them now show the same SEOBetter score, with GEO listed separately, and "Re-analyze" refreshes every surface.
• The test buttons on the Settings research tab no longer appear on sites where they cannot run. A short note explains why.

Generation quality fixes:

• Fixed: articles repeated the post title as the first heading, right above Key Takeaways.
• Fixed: listicles sometimes tacked on extra businesses that were not part of the verified, numbered list. Unverified extras are now removed.
• Fixed: generating several articles in a row could stall with a misleading "upgrade your hosting" error.

• Added Claude Sonnet 5 and Claude Fable 5 to the AI model picker, for both the Anthropic and OpenRouter providers.

• Fixed: article quality could drop when you generated more than one article in a short period. Later articles quietly fell back to a simpler writing engine. Every free article now uses the same tuned writing engine.
• Article outlines are now built by the tuned engine too, so headings and structure match the quality of the body text.

• Better references on free articles: research now finds more relevant sources, so articles get several on-topic reference links instead of one or none.
• Fixed: off-topic citations, such as a flood-monitoring source cited on a shopping article, no longer slip through.

• Local business listicles now render with numbered cards and a "list at a glance" index, the same as other listicles.
• Fixed: a false warning that flagged normal informational sections as unverified businesses. Genuinely made-up business names are still caught and removed.

Pre-launch cleanup batch:

• Removed random placeholder images. Articles now use your AI images, Pexels or Openverse photos, or no image at all. Never an unrelated placeholder.
• Clearer messages when a language or option belongs to Pro.
• Smaller plugin download.

• Honest article counters: the plugin now shows your real allowance (2 Cloud articles plus 30 articles with your own AI key, lifetime per site) instead of "unlimited".
• Free articles now fetch their writing guidance from the SEOBetter service, so free output follows the same SEO and GEO principles as the paid plans.
• The author bio box and author schema moved to Pro.
• Content type counts corrected to 17 across the plugin.

• First public build of the free plugin, after a full launch audit: accurate plan descriptions throughout, a free Cloud meter that correctly shows a lifetime allowance rather than a monthly one, and local business (Places) research confirmed as a free feature.