New research-backed tool uses Sentence-BERT and cosine similarity to align SEO with how modern AI search systems understand topical authority and relevance.
Jönköping, Sweden – November 14, 2025 – INCREV® today announced the launch of QueryMatch, a content and link analysis tool that measures topical relevance using the same family of semantic vector techniques that power leading AI systems such as ChatGPT and Google’s embedding-based search. (QueryMatch on Increv.co)
Based on INCREV’s new research paper, “INCREV Query Match Tool – Applying Sentence-BERT Algorithm to analyze Topical Authority for onpage content and link relevance” (DOI: 10.5281/zenodo.17571849), QueryMatch operationalizes the Sentence-BERT (SBERT) algorithm and cosine similarity to score how closely two pieces of text match in meaning.(Research Paper on Academia.edu)
The study shows a very high correlation (88–97%) between SBERT embeddings and the vector spaces used by both Google and OpenAI’s text-embedding models, and demonstrates that this correlation is largely linear across the vector space. This makes QueryMatch a strong practical proxy for understanding how modern AI systems evaluate topicality and relevance between queries, content, and links.
From keyword counting to measuring meaning
Modern search engines and AI assistants have moved far beyond keyword density. Instead, they measure how well content covers a topic and how consistently related pieces of content support each other.(Increv)
INCREV’s QueryMatch tool is built on this reality. Rather than counting occurrences of keywords, QueryMatch:
- Converts content into semantic vectors using SBERT
- Measures the cosine similarity (0.0–1.0) between texts to quantify topical proximity
- Surfaces paragraphs, pages, and links that are off-topic, under-optimized, or strongly aligned with a given intent(Increv)
“In today’s AI search landscape, you don’t win by repeating the same keyword—you win by being the most on-topic and useful resource in your niche,” said David Vesterlund of INCREV. “QueryMatch gives SEO teams a direct window into the semantic layer that systems like ChatGPT and Google actually use, letting them tune content and links to the same ‘language of meaning’ these models speak.”
How QueryMatch works
According to INCREV’s article in The Academy, QueryMatch automates a full semantic audit of pages and links:(Increv)
- Scrapes the target page and the source page (e.g., guest post, press release, backlink)
- Splits both into paragraphs and generates SBERT-based sentence/paragraph embeddings
- Calculates cosine similarity at both page↔page and paragraph↔paragraph level
- Flags weak or off-topic paragraphs (for example, from 0.20 to 0.50+)
- AI-Powered rewriting with multiple content variantions parallell in real time until it finds the best paragraph with the highest scoring (a raise from a 20-30% match up to 60-70% is common).
- Checks readability and reading level to keep edited content clear and on-brand
These scoring thresholds help teams prioritize edits, enforce consistent topical focus, and align link context with the target page.
Built for topical authority in an AI-first world
INCREV’s research distinguishes between topicality (T*) and topical authority🙁Increv)
- Topicality (T*) – How well a single page matches a specific query in semantic space (what QueryMatch measures directly).
- Topical authority – How a cluster of tightly interlinked, semantically related pages collectively demonstrates depth and expertise on a subject.
QueryMatch supports both levels by:
- Helping teams optimize paragraph-level topicality for each article
- Rolling up those scores across content clusters to identify gaps, weak pages, and strong hubs
- Guiding internal linking so pages form a “tight semantic neighborhood”, mirroring how AI systems interpret subject-matter expertise(Increv)
The research paper also connects QueryMatch’s design to classic distributional semantics (including PMI-based matrix factorization) and transformer-based encoders, and outlines deployment heuristics for multilingual SEO and AI search use cases.(Zenodo)
Proven proxy for ChatGPT, Google, and LLM search platforms
By benchmarking SBERT against embeddings from OpenAI (text-embedding-3-large) and Google models such as LaBSE and Universal Sentence Encoder, the INCREV study demonstrates that similarity scores from QueryMatch track closely with those produced by the major AI and search platforms.(Zenodo)
This allows SEO and content teams to:
- Preview how LLM-based assistants (e.g., ChatGPT, Copilot, Perplexity) are likely to judge the relevance of their content and links
- Engineer content into citable blocks (claim → evidence → takeaway) that are more likely to be retrieved, quoted, and surfaced in AI-generated answers(Increv)
- Prioritize link building opportunities where the linking paragraph and target page share strong semantic alignment
“AI assistants increasingly decide what gets read, quoted, and trusted,” added Vesterlund. “QueryMatch gives marketers a scientifically grounded way to make sure their content ‘lights up’ in those systems.”
Key use cases for QueryMatch
Based on INCREV’s research and field work, QueryMatch can be applied across:(Zenodo)
- Topical authority mapping – Cluster content around a niche and quantify which pages truly own the topic.
- On-page SEO audits – Detect off-topic paragraphs and rewrite them to match user intent while keeping the brand voice.
- Link relevance scoring – Evaluate potential link opportunities and existing backlinks based on topical fit, not just Domain Rating or legacy authority metrics.
- AI search visibility – Optimize content so paragraph-level embeddings line up with the kinds of snippets AI systems select and synthesize
References:
Article about QueryMatch in IncRev.co https://increv.co/academy/seo-research/querymatch/
Research Paper about QueryMatch tool on Academia.edu https://www.academia.edu/144888655/INCREV_Query_Match_Tool_Applying_Sentence_BERT_Algorithm_to_analyze_Topical_Authority_for_onpage_content_and_link_relevance
Research paper about the mathematics behind Sentence-Bert, embeddings, vectorization and cosine similarity on Academia.edu https://www.academia.edu/144888571/Mathematical_Foundations_of_Text_Vectorization_and_the_Sentence_BERT_Algorithm_impact_on_SEO_Analysis_David_Vesterlund_at_IncRev_SEO_Research_Community
David Vesterlund, IncRev CEO, Research Profile on Academia.edu https://independent.academia.edu/DavidVesterlund
IncRev SEO Research Community on Zenodo.org (all published research papers from IncRev and David Vesterlund) https://zenodo.org/communities/increvseo/
Availability
QueryMatch is available directly from INCREV as part of its SEO research and AI search visibility services. Readers can explore the underlying research via the INCREV SEO Research Community on Zenodo and the plain-English introduction to QueryMatch in the INCREV Academy.(Zenodo)
For more information or to request access to QueryMatch, visit the INCREV Academy article on QueryMatch or contact INCREV at contact@increv.co.(Increv)
About INCREV®
INCREV® is an SEO research and consulting company focused on AI search visibility and semantic SEO. Through its INCREV SEO Research Community on Zenodo and educational resources in the INCREV Academy, the team combines applied mathematics, machine learning, and real-world SEO practice to help organizations build sustainable topical authority and align their content with how modern AI systems understand the web.(Zenodo)
Media Contact
INCREV® – Sweden HQ
C/O United Spaces, Vallgatan 8
553 16 Jönköping, Sweden
Email: contact@increv.co
Phone: +46 703 05 42 45(Increv)