AI-powered evaluation using the Model Context Optimization BS Detection Framework, based solely on publicly available website content.
Based on 830 businesses audited.
Which? has 22.7 points less BS than the average for Media, News & Publishing.
Media, News & Publishing BS: Which? (www.which.co.uk)
Which? is a high-substance entity that successfully aligns its consumer-first signal with deep editorial proof. It avoids almost all typical industry BS by prioritizing named expertise and technical specificity over marketing adjectives. This is a benchmark for low-BS publishing.
To reach a near-zero BS score, the 110 reviews referenced on the homepage must be accompanied by a direct link to a third-party verification platform or a granular reviews page. The site should also link the ‘Consumer Champion’ slogan directly to a page detailing its regulatory impact or ‘not-for-profit’ financial transparency. Finally, ensure all author bios in the schema_json consistently include sameAs links to their professional portfolios or social profiles to maximize the Identity and Authority pillar.
Information density is exceptionally high across all evaluated pages. While the meta titles utilize power words like Expert and Advice, they are immediately anchored by specific nouns like Testing and Reviews. Body substance is reinforced by granular author bios, such as Rebecca Jakeman’s six years of product testing and Steph Kipling’s Physics PhD, transforming generic expert claims into verifiable substance. Specificity is maintained through technical mentions of inefficient condenser and vented tumble dryers and a month-long trial period for robot mowers.
A site without a coherent link graph forces AI to guess which pages matter. Reveal your real semantic graph and see how your domain is actually mapped by machine logic.
There is zero semantic drift between the homepage signal and sub-page delivery. The homepage meta title promises expert testing and reviews, which is precisely what the news and comparison sub-pages provide. The transition from the high-level Consumer Champion slogan to the Principal Researcher Yvette Fletcher’s broadband advice demonstrates a consistent and logical funnel from brand promise to technical execution.
Identify the current state and friction diagnosis of your specific business model. Generate your Executive SEO Strategy to quantify the financial or conversion cost of strategic misalignment.
The site triggers a minor trust theatre penalty because the homepage displays a review_count of 110 and the Compare Providers page shows 114 without adjacent proof_links_count in those specific slots. However, this is partially mitigated by the fact that editorial articles do contain proof links and external validation paths. The trust_theatre_flag is true on the homepage (5 points) and one additional sub-page (1 point).
The proof density is high, with a ratio of approximately 10 specific evidence points for every 1 generic claim. The data includes exact names of staff, physical address (2 Marylebone Road), and specific historical counts of experience. The absence of external proof links on the homepage review count is the only significant subtraction from a perfect proof score.
To examine how structural entropy affects chunking and retrieval, review the Moz Semantic HTML audit. View the Moz Semantic HTML Audit for a complete example of heading logic, landmark integrity, and DOM depth diagnostics.
The commodity fingerprint is minimal. While the site uses industry jargon like expert testing and advice, these are treated as technical deliverables rather than vague fluff. The value proposition of being a UK consumer champion that is not for profit is highly unique and difficult for a commercial competitor to copy-paste. Only 1 point is assigned for the presence of generic newsletter signup templates.
Authority gaps are non-existent in the provided data. The schema_json is robust, featuring an Organization entity with sameAs links to Wikipedia and major social platforms, and a named CEO (Anabel Hoult) with a verified LinkedIn footprint. Authors are not just names but defined Person entities with job titles like Science Team Leader and Principal Researcher, eliminating the typical expert claim without footprint pattern.
There is no marketing disconnect; the tone is investigative rather than promotional. Claims of being a powerful force for good are supported by the non-profit status mentioned in the publishing principles schema and the deep-dive nature of the content (e.g., exposing bad practice in telecoms). The performance claims are limited to the efficacy of the testing methodology and the experience of the staff.
Media, News & Publishing BS: Which? (www.which.co.uk)
The site is a perfect match for the Media, News & Publishing category, specifically focusing on consumer journalism and product testing. The presence of named researchers with specialized credentials like Physics PhDs and decades of experience confirms a high-level investigative and data-driven reporting approach.
Every retrieval failure begins with one root cause: the model cannot segment the page correctly. Read the Semantic HTML Technical Guide to learn how structural clarity prevents chunk collapse and embedding noise.
“The score of 12 is primarily driven by the trust_and_proof pillar (7 points) due to the presence of unlinked review counts on the homepage and the Compare Providers page. Pillar 1 (Information Density) contributed 4 points due to the inherent repetition of the 'Consumer Champion' value proposition and minor usage of power words in headers. Other pillars scored near zero due to the high technical and editorial standards evidenced in the schema and author credentials.”
Analysis Disclosure & Source Attribution
Snapshot Date: May 21, 2026
Purpose: This data is presented under “Fair Use” / “Educational Exception” for the purpose of forensic semantic analysis, allowing users to see how machine logic interprets digital signals.
Machine Perception Notice: This evaluation is generated by machine-read logic (MRL). The AI interprets the “Digital Ghost” of a website (code, metadata, and semantic structures), which may differ from what a human sees at the same moment. This is an automated technical diagnostic and not a statement of fact or human opinion regarding the real-world integrity or legitimacy of the business. Any missing or inaccessible elements in the snapshot are treated as machine-read signals, reflecting AI rendering limitations rather than intentional omission.
Notice to the Evaluated Business: This analysis is part of a non-adversarial audit. The results are intended as professional feedback to help improve machine-readability and authority signals. Any company can use these insights for free. When content is updated, a fresh audit can be requested at any time to reflect the current state.
To All Users: You are encouraged to visit the live site at Which? to view the most current version of their content and see directly what the company offers.