Feb 2026 SERPrecon update:

We just launched some major logic and backend updates to SERPrecon. This post will explain them in detail and what changes to metrics you might expect as a result.

• Updated Markdown formatting and Markdown editor on the optimize page

  • We made some changes to the markdown editor and the formatting to make it easier to edit/optimize and re-score your content. Markdown should now be cleaner, and the editor should be way more reactive.

• Multiple Language Support!

  • SERPrecon now fully supports English, Spanish, French, German, Italian, Portuguese, across all metrics.

  • Partial support for Russian, Japanese, Korean, Chinese_simplified, Chinese_traditional (BM25 and bigram NPMI may be less than accurate)

  • Other languages are partially supported.

  • If you want to try running a report against a language other than the ones listed, just select the closest language to it or "English."

• Updated Cosine Similarity & Semantic Relevance Logic

  • We've re-designed our semantic relevance logic to use a "chunking" method based on Google's colBERT model. For this reason, you may see "higher" scores across the board when compared to historical data.

  • We now analyze all content. We used to cut off content on extremely long pages.

• Improved Keyword Extraction

  • we no longer extract keywords if the page scores too low on semantic relevance. Since ranking results should always score above the threshold, this stops us from extracting keywords from server error pages and other random things that sometimes pop up.

  • Instead of extracting the top 25 keywords per page, we now extract the top 5 keywords per "chunk" and then return the overall top 25 keywords across all the chunks. This should provide more relevant keywords to the query.

  • We also ignore any chunk that scores as "irrelevant" as well as the bottom 20% of chunks from keyword extraction. This helps remove boilerplate, disclaimers, legal notices, etc from the data.

• Improved cookie consent logic. We got better at detecting and ignoring this type of content for analysis purposes.
  

We hope you find these improvements useful, and as always thank you for using SERPrecon. If you have any case studies of using our tool to dominate AI or SERPs, please reach out and share them - we'd love to feature your success on our website and socials.