Do AI recommendations for wedding vendors change from month to month?

Short answer: yes, more than most people expect. In our measurements the total barely moved between September and October 2026, but underneath it, about one in three positions changed. A business could lose its place on one AI service and win a new one on another in the same month, and end up with almost the same score.

This comes from our own data, not a survey. We ask the same questions to ChatGPT, Perplexity, Gemini and Claude, three times each, and save every answer together with the links the service showed. The method page explains the setup.

What we compared

Two wedding photography businesses in Sweden were measured on 12 and 14 September 2026 and again on 11 October 2026. Each had exactly the same 20 questions both times, written the way couples ask: broad questions, price questions, style questions and questions about specific venues. Every question went three times to each of the four services. That gives 160 question and service pairs, and 480 answers per month.

For each pair we sort the result the same way every time. Named reliably means the business was named, or its website cited, in at least two of three runs. Named sometimes means it happened in exactly one run. Not named means it never happened.

The total hardly moved

If you only looked at the bottom line, nothing happened: 65 reliable positions in September, 67 in October. Now look at the rows. ChatGPT went from 4 to 16. Perplexity went from 26 to 13. Claude went up, Gemini went down a little.

Perplexity dropped for both businesses in the same month. When two unrelated websites lose ground on the same service at the same time, the likeliest explanation is a change on the side of that service, not something either business did. We cannot see inside Perplexity, so that stays a likely explanation rather than a fact.

One in three positions changed

The table above counts positions. The more revealing view follows each of the 160 pairs from September to October and asks: did the result stay the same?

So a third of the picture changed in one month, in both directions at once. That is why a single number per month can mislead you. A steady total can hide large moves that cancel each other out, and the moves are where the useful information is: which questions you lost, which you won, and on which service.

Within a month, the runs mostly agree

This is not the same as answers changing every time you ask. Inside each measurement, the three runs gave the same result, either all three named the business or none did, for 258 of 320 pairs. The other 62 pairs split between runs.

So the runs done on the same day mostly agree, and the bigger change happens between months. The two habits solve different problems. Asking each question three times protects you from a lucky or unlucky single answer. Measuring every month shows how the answers move over time. You need both to know where you stand.

Named sometimes is worth watching

The small group of 12 pairs that were named sometimes in September behaved differently from the rest. Seven of those 12 were named reliably in October. Among the 83 pairs where the business was not named at all in September, 17 became reliable.

Seven of 12 against 17 of 83. The numbers are small, so treat it as a pattern to keep an eye on, not a rule. But it fits what we see in reports: a question where you show up in one of three runs is usually closer to a real position than a question where you never show up.

What this means for you

1. Do not judge your visibility from one answer, or from one month. One answer is a snapshot. One month is a better snapshot, but still a snapshot.
2. Look at each service on its own. In our data ChatGPT and Perplexity moved in opposite directions in the same month. A total across services hides that.
3. Keep the questions the same. A trend only means something if the questions do not change. If you rewrite them, you start a new measurement.
4. Look at what changed, not only the score. Losing a venue question on one service tells you something specific. A score that went from 65 to 67 tells you almost nothing.

This is how the full report works: twenty questions written for your business, three times each, to all four services, every month, with the same questions month after month so a change is a real change. If you want a first look, the free check asks four questions to ChatGPT and Perplexity in about a minute.

What this data cannot tell you

This is two businesses, one country, one kind of wedding vendor and two months. It shows that positions can move a lot while the total stays put. It does not show how much they move for every vendor, in every market, every month. A venue in the UK or a florist in the US could see something quite different.

It also cannot tell you why a service changed its answers. AI services update what they read and how they answer without announcing it, and from the outside we only see the result. That is the reason we keep measuring the same questions instead of trusting a single month.

We measure what AI services say. We do not, and nobody can, make them say something.