Holiday Park AI Decisionfrom AIVO, Inc. · when AI plans the break, does it choose your park?
Example Decision Test report · holiday parks

This is what a full Decision Test tells a park

A real test from our October 2026 research on UK holiday parks: AI agents on ChatGPT, Gemini, Perplexity each planned four kinds of stay 5 times per assistant: a family holiday, a touring pitch, a glamping break and buying a holiday home. 60 decisions in all. The park, its group, its rivals and the places are renamed; everything else is exactly as the agents recorded it.

Your AI shopping result · 60 runs · 3 assistants · 4 customer briefs
9 of 60

AI picked Park A outright in 9 of 60 shopping runs.

You were found in 43 of 60 runs but lost most of them before the pick. Sister parks in your group won 9 runs, as many as you. You do best with Glamping.

Found
43 / 60
Shortlisted
30 / 60
Picked outright
9 / 60
Win Rate range
8.9–24.1%
Confidence
Moderate: 5 runs per combination. Treat the rates as direction, not precision. 10+ runs gives firmer figures.
00 · Can AI agents reach you?

Agent access check

What a legitimate AI shopping agent could and couldn't read during this test. Blocks are recorded, never worked around.

Your pageVisitsStatus
park-a.example/our-parks/park-a17 Readable
park-a.example/our-parks/park-a/holidays9 Readable
park-a.example/our-parks/park-a/holidays/glamping-pods-park-a8 Readable
park-a.example/our-parks/park-a/ownership2 Readable
park-a.example/our-parks/park-a/touring-motorhomes1 Didn't load

Other sites that turned the agents away (they then relied on other sources): a pitch-booking platform (17), another national operator's site (12), a rival park's site (7), a local pub's site (2), a rival park's site (1), a property portal (1), a review site (1).

Decision chain · where your lost bookings broke down

51 of 60 decisions lost. The first barrier: discovery (28)

Every decision passes through the same six stages. Each lost decision is counted once, at the first stage where you fell out. Stages are in order, so a lost decision never reaches the stages after it.

1 · Discovery
Were you considered at all?
17 never surfaced in the agent's searches or pages; 11 found, but never put on the shortlist (no reason given)
28
2 · Access
Could the agent reach your pages?
4
3 · Decisive facts
Could it read the price, fees, availability and rules?
Facts the agents could not verify: price (7), rules (6), availability (5), fees (2)
8
4 · Evidence & trust
Could it rely on what it read?
Not yet measurable. 12 decisions stopped earlier, at access or readability, so the agents never got far enough to judge whether they could rely on your facts. Fix those, re-test, and this stage becomes visible.
0
5 · Buyer fit
Do you meet this customer's needs?
1
6 · Final choice
Considered, but something else won
5 lost to a rival the agent could verify; 3 lost to a rival judged a better fit; 2 lost to a sister business in the same group
10
Won9

Based on the agents' recorded searches, the reasons they gave when they cut you, and why they chose the winner. Their reasons are what they reported, not a window into how the model works. Fixing an early barrier often exposes the next one: re-test to see it.

Fix list · tied to the bookings it cost

What to fix, in order of the decisions it cost you

1 · Be findable for the requests guests actually make 17 decisions
Agents never came across your park. See what they searched for instead, and make sure your own pages and your listings answer those requests. The agents searched for, among others: “the bay heated glamping pods dog friendly pub walks prices November 2026”, “the bay the national park holiday parks lodges views restaurant walks 2027”, “the bay the river valley fully serviced touring caravan pitches dogs prices”.
2 · Publish dated prices on a normal page, beside the booking button 7 decisions
Agents could not verify your price for the dates. A table of season or date bands with nightly or weekly prices makes it checkable.
3 · Publish your letting, dog and site rules on the page they apply to 6 decisions
Agents could not verify the rules the guest or buyer asked about.
4 · Say availability in words, not only in a calendar 5 decisions
Agents could not confirm availability for the dates. A line such as 'May half-term: limited availability' is readable.
5 · Give agents the evidence the winners gave them 5 decisions
You were in contention and lost to a rival whose facts the agent could confirm. Publish the same kind of evidence: dated prices, availability, specifics.
6 · Let identified AI agents through your bot protection 4 decisions
Agents were turned away from your pages or listings. Ask your host or CDN to allow named AI agents on public pages, and check your listings on booking platforms.
7 · Review your offer for this request 3 decisions
Agents judged a rival a better match on features or location. This may be your offer rather than your website.
8 · Publish the annual site fee range, season length and other running costs 2 decisions
Buyers' agents could not verify your running costs. Put them on your ownership page.
9 · Decide which of your group's parks should win which request 2 decisions
A sister park in your own group was chosen instead. Point each kind of request to the park you want to win it.
Note · Genuine mismatches: not a website fix 1 decision
Agents read your facts and they did not meet the request. Worth knowing; nothing to fix online.

A decision can point to more than one fix; each is counted under the barrier it hit first.

01 · The journey

Where you dropped out

Every run passes through the same stages. This is how many of the 60 reached each one with you still in it.

Foundagent came across you
72%
43 / 60
▼ lost 13 here
Shortlistedmade the agent's list
50%
30 / 60
▼ lost 7 here
Comparedin the final head-to-head
38%
23 / 60
▼ lost 14 here · biggest drop
Pickedthe agent's recommendation
15%
9 / 60
02 · Every outcome

How all 60 runs ended

Not every run ends in one winner. We count every outcome.

9
47
4
Picked you9
Picked a competitor47
Picked nobody4
Stopped without deciding0
Didn't finish0
03 · Who beat you

Who AI chose instead

The 47 runs a competitor won outright.

Sister parks in your group
9
Rival M
8
Rival L
7
Rival H
4
Rival R
4
Rival K
3
Others (10 more, Rival G, Pub pods P, Rival N, Caravan Park, Rival O, Rival F, Rival W, Rival V, Rival T, Rival B)
12

Named as the place or brand the agent chose; an operator, group or listing site is folded into the place itself.

04 · By assistant

Each AI behaves differently

AssistantFoundPickedWinner most often
ChatGPT17 / 206Rival H
Gemini16 / 202Rival M
Perplexity10 / 201Rival L
05 · By customer

Where AI thinks you fit

Customer briefFoundPicked
Holiday homes for sale7 / 152
Glamping13 / 154
Holidays (caravan & lodge hire)11 / 153
Touring & motorhomes12 / 150

Holidays (caravan & lodge hire): “We're a family of four looking for a holiday park around the bay or the edge of the national park for the May half-term week in 2027. We'd like a caravan or lodge with nice views, somewhere to eat on site and good walks nearby, and our budget is up to £1,200 for the week. Which park would you recommend?”

Touring & motorhomes: “We have a touring caravan and want a fully serviced pitch around the bay or the river valley for the early May bank holiday weekend in 2027, with hook-up, water and drainage, dogs welcome, and ideally under £45 a night. Which park would you recommend?”

Glamping: “We'd like a two-night glamping break for two around the bay in early November 2026: a heated pod with a proper bed, ideally dog-friendly, near good walks and a pub, for around £150 in total. Where would you recommend?”

Holiday homes for sale: “We're looking to buy a holiday home caravan near the bay or the southern the national park, with a budget of up to [price], on a park that's open all year. We need to know the annual site fees and the rules on use and subletting. Which park would you recommend?”

06 · Why you lost

What the agents looked for and couldn't find

Taken from what the agents actually did: the pages they opened, what they searched for, and what they couldn't confirm. Observed means it happened in the recorded runs. Our read is our interpretation.

1 · Park A won when pod specs and price were visible
ObservedPark A was picked in 9 of 60 runs when agents could cite specific details: heated glamping pods, double bed, en-suite, pet policy, and the bay coastal walks, as in ChatGPT and Gemini Holidays and Glamping picks.
Evidence cited: 9 runs that picked you · Proof
FixKeep expanding this pattern: publish full spec sheets with price per unit type on every room/pitch page. · content
2 · Pricing pages blocked, so Park A often dropped at final step
ObservedPark A reached shortlist in 30 of 60 runs but was picked in only 9; multiple runs cite booking/availability pages as blocked or not showing date-specific price for short stays, caravans and pods.
Evidence cited: 8 runs where you were dropped · Facts
FixPublish crawlable per-night and per-date prices outside calendar widgets or cookie-gated booking tools. · web team
3 · Rivals with explicit nightly rates beat Park A
ObservedRival M and Rival L were picked over Park A repeatedly (Rival M 8, Rival L 7 across all runs) by showing clear published £/night figures where Park A pricing was hidden behind booking engines.
Evidence cited: 7 rival picks · Competition
FixMatch rival pages by listing a clear from-price per pitch/pod/caravan type near the top of each page. · content
4 · Gemini shortlists often but rarely picks Park A
Our readGemini shortlisted Park A in 9 of 20 runs but picked it in only 2, losing most often to Rival M (5) and Rival R (4), which offered clearer utility and fee specifics in the same runs.
Evidence cited: 5 rival picks · Competition
FixMatch the fee and utility transparency rivals show, especially annual pitch fee and subletting rules on ownership pages. · content
5 · Perplexity surfaces Park A least often of the three
ObservedPerplexity found Park A in 10 of 20 runs, versus 17 of 20 for ChatGPT and 16 of 20 for Gemini, reducing its chances to shortlist or pick compared with the other assistants.
Evidence cited: 3 rival picks · Discovery
FixImprove presence on sources Perplexity favours (review aggregators, structured listings) to raise discovery odds. · PR & reviews

These are the decision points we saw in this test. Fixing them makes a win more likely but doesn't guarantee one. Rerun the same test afterwards to see what changed.

07 · The 17 runs that never found you

What the agents searched for instead

Agent searchedTimes
“the bay heated glamping pods dog friendly pub walks prices November 2026”2
“the bay the national park holiday parks lodges views restaurant walks 2027”2
“the bay the river valley fully serviced touring caravan pitches dogs prices”2
“the bay holiday caravan sale 12 month season site fees subletting”1
“southern the national park holiday caravan sale open all year site fees subletting”1
“holiday homes for sale the seaside town 12 month season annual site fees”1
“365 days" "subletting allowed" static caravan the county site fees”1
“Rival D holiday home subletting rules”1
08 · Replay a run

ChatGPT · Holidays (caravan & lodge hire) lost to Rival L

01Search“the bay holiday park restaurant sea views the national park walks”
02Search“the national park edge holiday park restaurant lodge walks nearby”
03Search“May half term 2027 caravan holiday park the bay”
04Openpark-a.example/our-parks/park-a/holidays/
05Opena national operator's site
06Opena rival operator's site
07Openanother national operator's site (turned away)
08Shortlista national operator, Group A, Rival P
09Search“site:a national operator's site Rival L 29 May 2027 7 nights price”
Want this for your park?
Who AI agents choose instead of you, why, where each lost booking broke down, and what to fix first. £4.00 per agent run: one journey (3 assistants × 5 runs) is about £65. Pay as you go.
Set up your Decision Test → or start with the free Agent-Ready Check