The most interesting result in the whole set belongs to a venue. A manor house in Sweden, measured in September 2026 on eighteen questions, each asked three times to Perplexity, ChatGPT with web search, Gemini and Claude.
It never once held a stable place. It also appeared twenty three times. Both of those are true, and together they describe a situation almost every venue I have measured is in.
0
questions where the venue held a stable place on any service
23 of 72
question and service pairs where it appeared in exactly one run of three
11
answers that cited the venue own website
Nearly visible, everywhere
Thirty two percent of the question and service pairs produced an appearance. Not one of them repeated across runs. Ask the same question again a minute later and the venue was gone.
| Service | Stable places | Appeared in one run of three | Cited the venue site |
|---|---|---|---|
| Perplexity | 0 of 18 | 10 | 10 answers |
| Gemini | 0 of 18 | 9 | 0 answers |
| Claude | 0 of 18 | 3 | 1 answer |
| ChatGPT | 0 of 18 | 1 | 0 answers |
Eighteen questions, three runs on each of four services, September 2026.
Perplexity is the interesting column. Ten appearances and ten citations of the venue own website, which means Perplexity was reading the site, using it, and still not naming the venue consistently. The raw material was there. The confidence was not.
Who took the place instead
A short and very stable list. Another manor house came back in nineteen answers, a hotel in thirteen, and four more places between five and eight each. The same handful, over and over, across all four services.
That is what a venue is competing with. Not every venue in the region, but the five or six the assistants can describe without hedging. Getting into that group is a question of being described, in text, in more than one place.
Venue answers are built differently
Photographers get named out of photographer directories. Venues get named out of a wider mix: regional wedding guides, planner and supplier lists, real wedding features naming where the day happened, and in the United Kingdom the big two directories.
When the same measurement asked how a venue gets recommended in the UK, the answers came back full of British venue lists. Hitched and Bridebook are load bearing there in a way that has no Swedish equivalent. This is why category advice copied from an American or British blog post will send a Nordic venue in the wrong direction.
Three things to fix first
Write the practical page nobody wants to write. Capacity, what a wedding costs, whether guests can sleep there, what is included, how far from the nearest city. Those are the words in the questions. A page of atmospheric photography with a booking form answers none of them.
Get named in other people pages. Every photographer, planner, florist and caterer who has worked at your venue is a potential sentence with your name in it. Ask them to name the venue in the text, not only in the photo caption. Supplier lists work in both directions.
Turn the near misses into positions. Take the questions where you appeared once out of three and write the page that answers exactly that question. Those are the ones where the assistant already half knows you, and they move first.
What a measurement shows you
Your own questions, four services, three runs each, which questions you hold, which you nearly hold, who takes the place instead, and the sources the answers were built from. You can see an example report or read how the measurement works.
Questions florists ask next
Why do AI assistants recommend other venues in my area and not mine?
Usually because the other venues are described in more places than yours. In the measurement the manor house that came back nineteen times was not obviously better, it was better covered. Venue answers are assembled out of directories, wedding features, planner pages and supplier lists, and the venue that appears in more of those is the one the assistant can describe with confidence.
Do wedding directories matter for venues?
They matter more for venues than for any other category, and which directory matters depends entirely on your country. Answers about British venues lean heavily on Hitched and Bridebook. Answers about the Swedish market do not mention either. Find out what the answers in your own market are built from before you spend anything.
We appear sometimes but not always. Is that good or bad?
It is the best place to be short of a stable position, and it is where the smallest amount of work gives the biggest change. Appearing in one run out of three means the assistant knows you exist and does not reach for you reliably. Appearing in none means it has nothing to reach for. Those two need completely different fixes, which is why counting them separately matters.
How many questions should a venue measure?
Enough to cover the different reasons a couple picks a venue: the region, the type of building, capacity, whether you can stay overnight, and price. In this measurement eighteen questions across four services and three runs came to two hundred and sixteen answers, which was enough to show a very clear pattern.