Search the site · Ask Zorva

Results appear as you type. Press Enter and Zorva answers from this site’s own pages.

Web

What Search Engines Couldn't See at Spring Haven

A Hendersonville wedding venue's site hid its photos from crawlers, blocked AI bots by accident and lost years of old links. What we found and how we fixed it.

May 5, 2026Updated October 14, 20266 min read
Zorva assembling a glowing website wireframe in mid-air (illustration)
TL;DR

Spring Haven Mansion is a historic wedding venue in Hendersonville, and its site looked fine to visitors while hiding things from machines. Its photo page showed crawlers two words, a Cloudflare setting no one had chosen turned AI crawlers away, and three older sites had left 126 addresses behind. Each fix made what a couple sees and what a search engine reads the same page.

Until October, anyone who opened the photo page at Spring Haven Mansion saw dozens of wedding pictures. A search engine opening the same page read two words: "Loading gallery…". The pictures were real, and so was the problem. They arrived by script after the page loaded, and a crawler that doesn't wait never saw them.

Spring Haven Mansion sits at the end of Spring Haven Court in Hendersonville, a few hundred yards from Old Hickory Lake. John and Polly Wagnon built the house in the early 1800s. John was Andrew Jackson's friend and physician, and the house shares design elements with the Hermitage. Today it hosts weddings, elopements and corporate gatherings. We told the story of its design in our write-up of the build. This is about the less visible work: the places where the site said one thing to a couple and another to a machine.

Why couldn't search engines see the venue's photos?

Search engines couldn't see the photos because the public gallery page filled itself in by script, while the side the owners use worked fine. The owners manage their own gallery. They upload pictures through a private admin page, the files go to Cloudflare storage and a small database keeps the order. That part works well, and we didn't want to take it from them. The trouble was the public page, which started empty and filled itself in the browser. On top of that, every picture carried the same alt text, "Spring Haven Mansion wedding", so even a crawler that ran the script learned nothing from ninety-one images.

In October we moved the gallery into the HTML. The build writes all 91 pictures into the page, each with alt text describing what's actually in it, and that text doubles as the caption. When a visitor asks for the page, a small server function swaps in the owners' current list from the database, so a new upload still appears without a rebuild. The page also gained a two-paragraph introduction and four questions of its own. Couples come to a venue's site for the pictures, and a crawler should find them too.

Why was Cloudflare blocking AI crawlers from the site?

Cloudflare's own onboarding had switched on AI bot blocking, and in late September a routine check found Cloudflare answering GPTBot and ClaudeBot with a 403, while a normal browser and PerplexityBot got the page. The site's robots.txt allowed everyone. Nobody had decided to block anything.

The account's audit log explained it. The zone was added to Cloudflare on March 27, and one second later the onboarding switched on AI bot blocking. So from before launch, GPTBot, ClaudeBot, CCBot and Bytespider had all been turned away, in a setting the dashboard didn't make obvious. Our own scanner hadn't caught it either, because its check read robots.txt and nothing more. We changed the setting to match the other 72 zones in the account, and all four now get the page.

Whether a venue should let AI training crawlers in is a real choice, and some owners will say no. It just ought to be a choice. If your site is on Cloudflare, fetch it with a crawler's user agent and see what comes back, because your robots.txt may not be the whole story.

Twenty years of old addresses

Spring Haven has had three websites before this one: a static site from 2006 to 2013, a WordPress site from 2014 to 2019, and another WordPress site from 2024 until early 2026. The old pages were gone, but people's bookmarks, old blog links and Google's own memory of them weren't. Many of those addresses answered with a 404.

With the old host offline, we couldn't crawl it, so we built the list from the Wayback Machine and Search Console. That turned up 126 old URLs, and 117 of them now redirect to the page that answers them today. The old pricing page goes to the packages page, a post about the house as an event venue goes to the About page, and three photos people had found through image search point to the same files. One discovery was our own doing. Before the site had a 404 page, Cloudflare Pages served the home page for any unknown address, and that page's relative links had created 13 junk URLs that Google kept on file. They redirect now too.

What should a venue's llms.txt file say?

An llms.txt file should say only what the site's own pages say. This site carries one, a proposed standard for giving language models a plain summary of a site. No major engine has confirmed it reads one, so it's cheap insurance at best. Still, it shouldn't say what the site doesn't. Ours did. It mentioned tents, micro-weddings for 2 to 30 guests, drive times, the distance to the airport, a bridal suite, an in-house coordinator and single-day bookings, none of which appeared on any page.

In October we rewrote it from the pages alone, each line naming the page it came from. If a venue's AI summary promises a bridal suite the venue doesn't advertise, the couple who arrives expecting one is the one who finds out.

A venue is a place in the markup

The structured data needed the same honesty. It had a placeholder Person named "Spring Haven Mansion Owner", which we removed, so the markup names no one the site doesn't. The business is now typed as both an Organization and an EventVenue. That pairing matters, because EventVenue sits under Place and CivicStructure at schema.org, not under LocalBusiness, so on its own it says what the building is and nothing about the business renting it. Each package offer now carries its own URL, so a price in the markup points to the page that shows it.

The country gate that stopped PageSpeed

The last surprise was a firewall rule. In April someone had added a Cloudflare rule blocking every visitor outside the US and Canada. It looked harmless for a venue in Tennessee. But PageSpeed Insights runs Lighthouse from Google's own data centers, and on October 11 a run from Taiwan was refused and scored the site's SEO at 40. We deleted the rule and moved the gate into the site's own code, limited to the US, with search crawlers, Lighthouse and PageSpeed let through by user agent. A gate a crawler can walk past is one you can live with.

For anyone running a venue site

Open your photo page with scripts turned off. Fetch your home page as GPTBot. Compare your llms.txt with what your pages actually say. Look up your old site in the Wayback Machine and make sure every address it had still leads somewhere. And check whether a country rule is quietly refusing the tools that grade you.

If you'd like to know how the answer engines describe your venue today, our AI visibility check shows what they say and where it came from. The wedding venues page explains how we work with venues like Spring Haven.

FAQ

How do I market a wedding venue?

Start with a site that shows every photo, price and package in plain HTML, so search engines and couples read the same thing. Name your town in the page titles, keep your old web addresses redirecting, and give your photo page real descriptions. Spring Haven's titles now name Hendersonville, and its 91 gallery photos describe what's in them.

How do I advertise my wedding venue?

Use regular Google Ads aimed at your area, sent to a packages page with prices. Google's Local Services Ads don't list wedding venues among their US categories, so that shortcut isn't available. Make sure the landing page loads fast and shows the guest counts and hours each package includes.

Why aren't my venue photos showing up in Google?

Often because the gallery loads by script, so a crawler sees an empty page or a loading message. Spring Haven's photo page read "Loading gallery…" to crawlers until we wrote all 91 pictures into the HTML. Give each image alt text that says what it shows, since identical alt text tells Google nothing.

Should a wedding venue block AI crawlers?

That's the owner's decision, but make it on purpose. Spring Haven's Cloudflare zone had blocked GPTBot, ClaudeBot, CCBot and Bytespider since setup, though its robots.txt allowed them. If you want answer engines to describe your venue accurately, let their crawlers in and check what they actually receive.

What happens to my old venue website's links when I get a new site?

They break unless someone redirects them. Spring Haven had three earlier sites, so we rebuilt the list of old addresses from the Wayback Machine and Search Console. Of 126 old URLs, 117 now send visitors and search engines to the page that answers them today, including the old pricing page.

Ready to act?

Put any of this to work for you.

Send us the site. A written fix plan follows in one business day.

ReplyOne business day
ProposalWritten · scoped · priced
ContractMonth to month
Proposal request

Tell us what you need

No spam · One business day reply · Month to month