From 14 to 100: 5 tourism sites vs generative AI

In May 2026, 5 tourism businesses in the Castellane area (Gorges du Verdon, Alpes-de-Haute-Provence, France — the area covered by LiveVerdon) installed Livada SEO. Three weeks later, 3 of them score 100/100, grade A. The other 2 are at 80-87/B, with precise, actionable gaps.

This page publishes the real data, collected on June 9, 2026. It breaks down signal by signal what changed -- and why lodging schema and Wikidata sameAs are the two highest-impact levers.

Methodology note: Livada SEO's AI Visibility Score measures 11 signals specific to AI citation (ChatGPT, Perplexity, Gemini, Claude). This score did not exist before the plugin was installed -- it is the plugin that creates the measurement framework. The starting score (~14/100) is calculated retroactively by applying the algorithm to each site's initial data.

The 11 signals measured

Each signal is evidence that a site can be understood and cited by AI. The theoretical total is 108 pts; the score is displayed out of 100.

Signal	Max points	What it changes
llms.txt for AI crawlers	20	Structured map for AI to navigate the site
Knowledge Graph entity	12	Organization published, identifiable by name and type
Complete lodging schema	12	Address, GPS, phone → maximum local trust
Declared areas of expertise	10	AI knows what to cite you for
E-E-A-T signals	8	Founder + description = declared authority
Structured FAQ data	10	Q&A directly extractable by AI
Question-format headings	8	Surfaced as answers in conversational AI
Meta description coverage	6	A citable summary on every page
XML sitemap	6	Guaranteed discovery by every crawler
Snippet directives	8	AI can cite longer passages (max-snippet)
Wikidata sameAs entity	8	Off-site identity verifiable by every AI

The starting point -- before AI optimization

A tourism site running a standard SEO plugin (Yoast, SEOPress, or similar) typically covers two signals:

• Meta descriptions (6/6) -- standard SEO plugins handle these well.
• XML sitemap (6/6) -- generated automatically.

Everything else -- llms.txt, structured Knowledge Graph, lodging schema with GPS, Wikidata sameAs, expertise, E-E-A-T, FAQPage -- is missing. Starting score: ~12-14/100, grade F.

These 9 missing signals aren't cosmetic. They are exactly what ChatGPT, Perplexity and Gemini use to decide whether to cite a business in their answers.

Results as of June 9, 2026

Business	Type	Score before	Score after	Grade
Livada.fr (Castellane)	Local agency	~14/100	100/100	A
Hôtel du Roc (Castellane)	Hotel, 2-star	~14/100	100/100	A
Spa Castellane	Spa & massage	~14/100	100/100	A
Camping La Ferme (Castellane)	Campground, 3-star	~14/100	87/100	B
Entre Ciel et Verdon	Gîte / rental	~14/100	80/100	B

3 sites at 100/A in under 3 weeks. The 2 sites at 80-87/B have identified, actionable gaps (see next section).

The two highest-impact signals: lodging schema + sameAs

1. Complete lodging schema -- +12 points

The "complete business listing" check went from 0 to 12/12 across all 5 sites after configuring LodgingBusiness, Campground or LocalBusiness schema with:

• Full address (street, city, postal code, country)
• GPS coordinates (latitude / longitude)
• Phone number
• Official URL

Without these 4 data points in structured schema (JSON-LD), AI cannot locate you with certainty. It may have your address in its text, but it cannot link it to a known entity. That's the difference between "mentioned" and "identified".

2. Wikidata sameAs entity -- +8 points

All 5 businesses got a Wikidata entry created and sourced in June 2026. Livada SEO automatically injects the identifier into the sameAs field of the JSON-LD schema.

Business	Wikidata entity
Livada.fr	Q141325080
Hôtel du Roc Castellane	Q140007358
Spa Castellane	Q140007421
Camping La Ferme	Q140007381
Entre Ciel et Verdon	Q140007422

Result: when a user asks ChatGPT or Perplexity "best hotel Castellane Gorges du Verdon", the AI can now connect Hôtel du Roc to a verifiable, sourced entity in its knowledge graph -- not just one URL among many.

The signal-by-signal breakdown -- 3 sites at 100/A

Here is the full picture for Hôtel du Roc Castellane, representative of the 3 grade-A sites:

Signal	Score	Detail
llms.txt	20/20	Served at the domain root
Knowledge Graph entity	12/12	Organization published in JSON-LD
Complete lodging schema	12/12	Address + GPS + phone filled in
Areas of expertise	10/10	9 topics declared
E-E-A-T signals	8/8	Founder + organization description
Structured FAQs	10/10	3 pages with FAQPage schema
Question-format headings	8/8	7/12 pages sampled
Meta descriptions	6/6	12/12 pages covered
XML sitemap	6/6	Present and valid
Snippet directives	8/8	max-snippet active
Wikidata sameAs	8/8	Q140007358 declared
Total	108/108 → 100/100	Grade A

The gaps on the 2 sites at 80-87/B

Camping La Ferme -- 87/B

Missing signal	Current score	Potential gain	Action
Areas of expertise	5/10	+5	Declare 5+ topics under Settings → AEO
Structured FAQs	5/10	+5	Enable the FAQ schema on 2 more pages
Question-format headings	4/8	+4	Rewrite the H2/H3s on money pages as questions

Potential with these 3 fixes: 87 + 14 = 101 pts → 100/A.

Entre Ciel et Verdon -- 80/B

Missing signal	Current score	Potential gain	Action
E-E-A-T signals	4/8	+4	Add organization description + founder name
Structured FAQs	0/10	+10	Add a FAQ block to the main pages
Question-format headings	0/8	+8	Rewrite headings as visitor questions

Potential with these 3 fixes: 80 + 22 = 102 pts → 100/A.

What this actually changes

The AI Visibility score measures a site's technical readiness to be cited by AI -- not actual citation (which also depends on brand awareness and content volume). But technical signals are the prerequisite: without them, even the best content stays invisible.

What we observed across these 5 sites:

• Long-tail queries move: "family hotel Verdon", "spa Castellane massage", "campground with pool Castellane" -- long-tail phrases that generative AI pulls directly from structured FAQs.
• The Knowledge Graph fills in: once the Wikidata entity is declared and lodging schema is complete, some sites start appearing in Google knowledge panels (not guaranteed, but correlated).
• Perplexity's answers cite pages with FAQPage and complete meta descriptions more often. Queries like "things to do Verdon" and "what to do Castellane" surface pages from these sites.

Why these results in so little time?

Local tourism is a sector that is under-optimized for AI. The big operators (Booking, Tripadvisor, Airbnb) concentrate their optimization on their own aggregator pages. Independent businesses start from further behind -- which means the gains come proportionally faster.

Livada SEO is built for exactly this profile: an independent hotel, campground, spa or gîte, in a tourism region, with a limited budget but real local value. The plugin automates the 11 AI signals in a few minutes of setup -- no developer, no agency.

"We installed the plugin one evening. By the next day, the score was already at 72/B. Two weeks later, 100/A. We'd never seen that with any other SEO plugin."

— Florin Livada, founder, on configuring livada.fr

Reproducing these results on your site

All 11 signals are within reach of any tourism business that installs and configures Livada SEO:

1. Complete lodging schema: Settings → Schema → fill in address, GPS, phone.
2. Knowledge Graph entity: Settings → General → AEO section → describe the organization.
3. llms.txt: enabled by default on install.
4. Wikidata sameAs: create a Wikidata entry (or have Livada do it) then paste the Q identifier into Settings → General → AEO → Wikidata entity ID.
5. FAQPage + question headings: add a FAQ block to 3 money pages, toggle FAQPage on in the Livada SEO metabox, rewrite 5+ H2/H3 headings as questions.

Estimated time to reach 90+ (grade A): 45 minutes of setup for a business with the content already in place.