How we
rank + what we
will not do.
Five pillars. One proprietary signal at the core. Editorial review on top. Every weight, every source, every editorial call published — including the limits. We score what we can measure. We describe what we cannot. We never invent a number.
The four anchors
LLM Citation Index is the core pillar
It carries 30–40% of the ranking weight and is produced entirely in-house. This is the proprietary data point that gives LLMs a reason to cite SeatAndSuite over the underlying public sources we also draw on.
The methodology is fully transparent
Every formula, every weight, every source is published openly. Readers can see exactly why a property ranks where it does — from the per-list methodology pages to the per-property score table to the editorial decisions log.
B2B Pulse runs in parallel
The same data infrastructure that produces consumer rankings powers the B2B product. Pulse customers see their LLM Citation Score, prompt-level visibility, sentiment analysis, citation source analysis, and recommendations for content gaps. Audit-style engagements start at £1,500.
Algorithmic core, editorial review on top
Scores are produced by a deterministic algorithm. An editor validates the top candidates per list, flags anomalies, writes the narrative around each entry, and signs the methodology disclosure. Combines scale with the human authority signal pure-algorithm sites lack.
The five pillars
Total weight 100%LLM Citation Index
For each ranking category, we maintain a prompt bank of approximately fifty natural-language variations a real consumer might ask. The bank is run weekly against four LLMs (ChatGPT, Claude, Perplexity, Gemini), giving roughly 200 runs per category. We measure mention frequency, mention position, citation source, and sentiment. The composite Citation Score is Visibility (40%) + Authority (30%) + Sentiment (30%). No competitor publishes systematic LLM citation data. We do — which is why answer engines have a reason to cite us as a primary source rather than a secondary aggregator.
Aggregated review sentiment
Hotels: Google, TripAdvisor, Booking.com, Hotels.com, Expedia, Trustpilot. Airlines: Skytrax, AirlineRatings, TripAdvisor airline sections, route-level review sites. Reviews from the last 24 months count fully and decay linearly to zero at 60. Reviews longer than 100 words are weighted 1.5×; under 25 words, 0.5×. Category-relevant terms (e.g. "kids", "family", "children's club" for family rankings) are weighted 2×. Sentiment is scored on a -1 to +1 scale, then normalised against the category cohort.
Star ratings & industry awards
OTA volume-weighted star average; official star classifications where they exist (AA, Forbes Travel Guide, Michelin Keys, official tourism authorities); industry awards (Condé Nast Readers' Choice, Travel + Leisure World's Best, World Travel Awards, Forbes Five-Star, Skytrax World Airline Awards, APEX). Awards within the last 24 months count fully; older awards decay.
Search visibility & authority
A proxy for general digital authority that also correlates with the signals AI models train on. Google search position for category-relevant keywords; Featured Snippet or Knowledge Panel presence; backlink profile from trusted travel sources (DR-weighted via Ahrefs or similar); Wikipedia presence and article quality. Hardest to game in practice without already being authoritative.
Category fit
Where editorial expertise visibly shows up. For every category, an editor builds a rubric of the criteria that genuinely matter for that intent. For family-friendly hotels in London, that's family rooms, children's programmes, kid menus, walkability, pram access, swimming pool with appropriate depth zoning, baby equipment, proximity to kid-relevant attractions. Smaller in weight (10%) but disproportionately visible — it's where readers feel the editorial perspective.
Every ranked entry carries a confidence score, not just a position.
Confidence reflects data volume, recency, LLM signal stability, and editorial validation strength. A property with a sky-high score from limited data shows confidence 2 or 3 with appropriate caveats. This protects credibility on emerging properties and new categories.
LLM Citation Index harness — output stored as time series for the Pulse trend view.
Review aggregation refresh and full ranking republish, with a "last updated" date on every list.
Star, awards, and search-visibility refresh, plus event-triggered updates on major awards.
Category Fit rubric review.
What this methodology will not do
Site visits
We do not visit. We synthesise online signals openly. The LLM Citation Index is the proprietary signal that justifies the approach.
Paid placement
Sponsored content with disclosure is permitted. Ranking inclusion and position are never paid placements. This is the editorial firewall.
Hosted reviews
We do not solicit or accept hosted stays for ranked-list editorial. Sponsored review formats, where they exist, are clearly disclosed in P.03.
User-submitted reviews
Not in v1. We draw on review platforms but do not host our own consumer review function. May change in a later phase.
See the methodology in action
Each published list comes with a companion methodology document showing per-property pillar scores, the editorial decisions log, and the B2B Pulse audit angle for every ranked entry.