Public methodology
How LobbyBear scores your hotel
Everything that goes into your report, with every weight. These numbers are read straight from the code that computes your report, so this page and your report always agree. Last revised 2026-10-30.
Never a factor: who built or hosts your website, your website vendor, and whether you pay LobbyBear. Details
AI Visibility (0–100)
Each week we ask AI assistants the questions travellers really ask about your destination (for example “best quiet hotel in Santa Monica for couples”). For every answer we note whether your hotel is named, and in which position.
Each answer gets points by where you appear. Assistants usually name 7 or 8 hotels and travellers read the whole short list, so the top three places all earn full points:
| Input | Points |
|---|---|
| Named 1st | 1 |
| Named 2nd | 1 |
| Named 3rd | 1 |
| Named 4th | 0.88 |
| Named 5th | 0.76 |
| Named 6th | 0.64 |
| Named 7th | 0.52 |
| Named 8th | 0.4 |
| Named 9th | 0.28 |
| Named 10th | 0.25 |
| Named after position 10 | 0.25 |
| Not named | 0 |
Only questions that fit you count. A question about a price level (budget, mid-range, upscale, luxury) counts only for hotels at or next to that level, so a luxury hotel is not marked down for missing “cheap hotels in …” answers. Answers where the assistant refused or named no hotel at all are left out too, so they can’t pull you down.
From points to a score. We average the points over those answers, then turn the average into 0–100 on a curve that rises fastest at the start: you don’t have to top every answer in your area to score well. Being named in the first places in about a third of the answers that fit you is a good score.
| Input | AI Visibility |
|---|---|
| Average 0.05 | 21 |
| Average 0.1 | 37 |
| Average 0.2 | 60 |
| Average 0.3 | 75 |
| Average 0.45 | 88 |
| Average 0.7 | 97 |
Runs. AI answers vary, so we ask every destination question 5 times per assistant each week. Questions written specifically about your hotel (“probes”) are asked 3 times, in your primary market only.
Ranges. We compute a 90% interval by resampling the answers. When that interval is wider than 5 points we show your score as a range (for example 38–47) instead of one number. Anything based on the 3-run probes is always shown as a range.
Grade
Your AI Presence grade is your AI Visibility, put in one of three bands. The same bands apply to each part score.
Poor
0–39
Fair
40–69
Good
70–100
A grade changes only when the score crosses 40 or 70 by more than half its likely range; otherwise last week’s grade is kept, so noise near an edge does not flip it.
Part scores: every input and weight
Three part scores (each 0–100) explain the grade. A part score without enough data is shown as “Not enough data”, never as 0. “Your website” shows “Not checked” when we don’t have the hotel’s website address yet.
AI answers
Your AI Visibility: how often AI assistants name you in answers to travellers’ questions, weighted by position.
| Input | Weight |
|---|---|
| AI Visibility (share of answers naming you, position-weighted 1 / 0.8 / 0.6 / 0.45 / 0.3) | 100% |
Your website
Whether AI assistants can read and trust your own website.
| Input | Weight |
|---|---|
| AI assistants allowed to read the site (robots.txt) | 25% |
| Hotel details in a machine-readable block (schema.org Hotel/LodgingBusiness) | 20% |
| Key facts (price level, area, top features) written on the page | 20% |
| Answers to common guest questions (FAQ) | 15% |
| Site reachable over https and fast | 10% |
| Fact sheet for AI assistants (llms.txt) | 5% |
| Sitemap | 5% |
Listings & reviews
Whether booking sites and maps agree with your confirmed facts, and how guests rate you.
| Input | Weight |
|---|---|
| Listings that agree with your confirmed facts (share of checked listings) | 60% |
| Guest review rating (normalised to 0–100) | 40% |
Wrong facts
We only call something AI says about you “wrong” when both are true:
- it contradicts a fact you confirmed (valid on the day the answer was given), and
- the same assistant said it in at least 2 of 5 runs of a destination question, or at least 2 of 3 runs of a probe about your hotel.
One odd answer is never reported as a wrong fact.
Competitors
We suggest the hotels you really compete with using a competition score from 0 to 100. Each factor is scored 0–1 and weighted as below. If we lack data for a factor, it is left out and the other weights are scaled up so they still add to 100.
| Input | Weight |
|---|---|
| Named together by AIHow often AI assistants name the two hotels in the same answer. | 30 points |
| Same kind of guestsOverlap in who stays (couples, families, wellness, …). | 20 points |
| Similar priceOverlap of nightly price ranges. | 15 points |
| Class & styleSimilar star class and type of property. | 15 points |
| LocationHow close the two hotels are. | 10 points |
| AmenitiesShared features such as pool, spa, yoga, restaurant. | 10 points |
Direct competitor
80–100
Suggested
65–79
Weak match
50–64
Not a competitor
0–49
You can always add or remove competitors yourself.
Bookings at risk (estimate)
An estimate of the monthly bookings you may be missing because AI assistants recommend other hotels instead of you. How it’s worked out:
monthly demand for the questions where you trail your set × your gap to the set’s median hotel × share of planners who book (low / high) × your nightly rate × average nights per stay
Every input (all shown next to the estimate)
- Monthly demand for the questions where you trail your set (low end)
- Monthly demand for the questions where you trail your set (high end)
- Your gap to the median hotel in your set on those questions (share of answers)
- Your own average nightly rate (from live rates for your hotel)
- Currency of that rate
- Average length of stay in your destination (nights)
- Share of AI-assisted planners who book a recommended hotel (low end)
- Share of AI-assisted planners who book a recommended hotel (high end)
- Where the demand numbers come from (named next to every estimate)
- Demand is an estimate of how many people ask those questions each month for your destination. Its source is named next to every estimate.
- Booking share uses a range of 1% to 3% of AI-assisted planners — that’s why the result is always a range.
- Always a low–high range labelled “estimate”, never one number and never a promise of revenue.
- Hidden when any input is missing. If we don’t know your rate, demand or gap, we show nothing rather than guess.
AI assistants we ask
We ask ChatGPT, Perplexity, Claude and Gemini through their official APIs, with web search switched on — the same kind of answer a traveller gets.
Your scores count the AI engines switched on for your hotel. Each score says how many engines it is based on, and when that set changes the timeline marks it, because scores before and after are not directly comparable.
- ChatGPT
- Perplexity
- Claude
- Gemini
Google AI Overviews: Excluded until we have a licensed, approved data source. We do not scrape Google.
What we never use
None of these is ever an input to any score, grade, ranking or estimate:
- Who built or hosts your website
- Your website vendor, theme, page builder or booking engine brand
- Whether you pay LobbyBear, or for which plan
- Advertising you buy anywhere
Your website vendor is never a factor. We only check what AI assistants can read on your site (see “Your website” above) — the same checks whoever built or hosts it. Paying LobbyBear does not change your score or anyone else’s.
Versions
Every report records the method versions it was computed with. When a method changes, the version changes and this page is updated.
| Method | Version |
|---|---|
| AI Visibility scoring | sv2 |
| Competition weights | cw1 |
| Grade, part scores and bookings at risk | sm1 |
Last revised: 2026-10-30.