Wedding photographers and ChatGPT: the crawlers, and what couples ask

Which OpenAI crawlers matter to a wedding photographer?

Two of the three. OpenAI's crawler documentation names OAI-SearchBot, the AI crawler that indexes pages for ChatGPT search, and ChatGPT-User, which fetches a page when a user asks ChatGPT about it. The third, GPTBot, collects content that may be used to train OpenAI's models. The document draws the line plainly: a site that disallows OAI-SearchBot is removed from ChatGPT search answers, and a site that disallows GPTBot is not.

So there are two decisions, and they are separate. Whether OpenAI may train on your site's content is a licensing question, answered in the GPTBot group of your robots.txt. Whether ChatGPT search may cite your pages is a visibility question, and it turns on OAI-SearchBot. A file that blocks GPTBot and says nothing else leaves search where it was; robots.txt for AI crawlers reads the file token by token.

An answer can also come from what a model learned in training rather than from a fresh search, and then it reflects your site as it stood when it was collected. That is one way an answer can mention a package you retired or a studio name you dropped, and no robots.txt line reaches back to change it.

Checking that ChatGPT's crawlers can reach your wedding pages

1. Open /robots.txt on your domain and look for a group naming OAI-SearchBot. If there is none, OAI-SearchBot follows the * group, so read what that group disallows; a Disallow: / there shuts out every crawler without a group of its own.
2. Run the crawler access check. It reads the file the way each of the nineteen AI tokens would and reports which are allowed, the two OpenAI tokens that serve answers among them.
3. Fetch one wedding page as the crawler would and look for a sentence you know is on it: curl -sL -A "OAI-SearchBot" https://www.example.com/weddings/ | grep -c "your package name". A count of zero means the words are not in what came back: a challenge page, a script shell or an error. In our audit of 9,483 small business websites (40 trades, 90 metros, late August to early September 2026), 7.2% served a challenge page instead of the site, a block nobody wrote into robots.txt; that figure covers every trade, not photographers alone.
4. Read a refusal of your own test as a lead, not a verdict. Some bot protection checks whether a request that claims to be a known crawler comes from that operator's own addresses, so look in your CDN's bot settings for how OpenAI's crawlers are treated before concluding anything.
5. Open a gallery page with JavaScript turned off, or view its source. If the photographs, captions and text are missing, a fetcher that runs no scripts gets none of them; 3.0% of the audited sites delivered their content only through JavaScript.
6. Check where the galleries live. A portfolio on a gallery platform's own domain sits on a different host with its own robots.txt, written by the platform and not by you, so keep public wedding stories on your own site. Keep client proofing galleries behind a password: robots.txt asks crawlers to stay out, but Google's robots.txt documentation notes that a blocked URL can still be indexed when other pages link to it.

The five facts a couple's question turns on

A couple can put a date, a venue, a style, a budget and a deadline into one sentence: "documentary wedding photographer in toronto, free june 12, under [price], who delivers within [weeks]". An answer drawn from your pages can speak to each part only where a page states it.

Two of these need care. An availability line that goes stale is worse than none, so date it and change it the day a date is taken, or state only what does not change: how far ahead you book and how a date is held. And a venue list is only as good as the pages behind it; each venue should lead to a wedding story with words on it, the subject of the article on a readable portfolio.

Prices, travel and delivery have more to them than a sentence each: what moves a price, where a travel zone begins, what arrives in which form. The article on the pages beyond the portfolio takes them one at a time, and the photographers' hub lists every fact a client asks about.

Reading the sources ChatGPT gives for a couple's question

ChatGPT search answers can carry citations to the pages they drew on, and those pages are the most direct evidence you have of where a couple's question leads. Ask the whole question the way a couple would, with the date, the venue or the town, and the style, with search on, and open every source listed. For each one, write down three things:

• Whose page it is: a photographer's own site, a wedding directory, a venue's list of recommended vendors, a planning blog.
• For another photographer's page, what it states for this question that yours does not: a dated availability line, a starting price, the venue by name, travel terms.
• For a directory or a venue list, what it says about you, if you appear at all, and whether that matches your own site.

One question on one day is a sample, not a measurement. It shows you which pages to set beside yours; whether you are named across the questions that matter, reading after reading, is what a fixed panel records, run the way our methodology describes.

What OpenAI's documentation leaves unsaid

It describes which crawler does what and how a site owner allows or refuses each one. It says nothing about how ChatGPT chooses between the photographers it could name, and nothing on this page claims to know. What can be observed from outside is whether you are named, for which questions, beside whom and with which facts, by asking the same questions every month; AI visibility for photographers sets out a panel by genre and place.

Google's side uses other tokens. Googlebot crawls for Search, AI Overviews included, and Google-Extended controls whether crawled content is used for Gemini training and grounding, as Google's list of common crawlers sets out. AI search engines at a glance puts the assistants side by side.