What a stranger finds when looking for a church 

Almost nobody arrives at a directory by accident. A stranger with no history starts at Google or an AI assistant. The people who open a directory already know the name of the denomination: a family moving across the country, a pastor sending someone on, somebody raised in the tradition and looking for it again in a new town. They use the official directory because of the name they know and trust.

Across all three denominations, 367 of 3,651 listed links — one in ten — lead nowhere a visitor can use. Expired addresses. Pages removed. Sites that stopped answering years ago. Placeholder pages left by a hosting company after an account lapsed. The listing still looks authoritative, but it is no longer accurate. The ordinary name for this is link rotWhen links stop working because addresses change, sites move and accounts lapse. Nothing about the link looks broken until somebody clicks it.Read the full entry →.

Type a town, get a congregation, follow the link. Every part of that sequence has to hold for a directory to work, and the link is the part that drifts and breaks over time. Somebody at a church changed its website address and forgot to tell the denominational office three states away. Nothing on either end announces the break, and the only person who finds the problem is the visitor looking for a church.

Nothing requires a denomination or regional body to publish a directory of its churches. But whoever publishes one owns what it says.

We did not take the first pass as the answer. Automated checks can get refused for innocent reasons: one site turns away unfamiliar visitors, one answers too slowly to look alive, one has a security certificate that the test distrusts but browsers accept. For each apparent failure, we tried again in a different way, and about a third came back alive; those are not in the number above.

Some of these addresses answer perfectly well, but are not churches any more. A hosting company's "domain parkedA web address that still answers but has nothing of yours behind it: a hosting company’s placeholder, an advertising page, or a notice that the account has closed. Everything about it works. None of it is yours any more.Read the full entry →" placeholder. A page saying the account has been deactivated. An empty shell with nothing on it. Every naive automated check counts those as working sites, and we did too until we looked at them by hand. They are the failures that look like success: nothing is broken, and nothing is there.

Nothing about finding these was difficult. We followed links a directory already publishes, from outside, with no access to anything private, and most failures gave themselves away in under a second. No judgment was required to tell a dead address from a live one. The check is trivial. Nobody is running it.

1,048 of 4,699 congregations are listed in their denomination's directory with no website whatsoever. We hand-checked a sample of them: an entire region of one denomination and a random draw from the rest of the country.

About a third of the churches recorded as having no website do have one. Real ones. Sites on Church Center and Faithlife, a Korean-language congregation, one church with its own phone app. No one ever updated the directory, or whatever record it was generated from.

The two thirds that genuinely have none divide almost evenly. One has nothing findable at all. No web page. No directory listing. Nothing. Several of these congregations have been meeting for fifty years, and the web has no idea they are there. The other lives on Facebook, which gets a section of its own, below.

Pew Research Center asked Americans who had looked for a new congregation how they went about it. In 2016, 37% had searched online — 47% among those whose search was within the previous five years, and 59% among adults under thirty. That is the most recent citable data we could find, but the rate was increasing during Pew's own study period.

A congregation with no website is not necessarily hidden or failing. It is reachable by anyone who already knows it exists, and for a church that grows through word of mouth in a tight-knit community, that may cost very little. For a church hoping to be found by someone new in town, having no website is a problem.

No one publishes how many American congregations have a website. It would not be hard to count, but no one ever has. There is one census big enough to have answered this: Faith Communities Today put 180 questions to 15,278 congregations in 2020. It asked about livestreaming and online giving, but not about websites. A website was not worth one of the 180.

Our own figure of 1,048 congregations, or a little over a fifth of the total, has nothing to compare it against, so we cannot tell you whether these three denominations are typical. Further study of more denominations would provide the answer.

As far as we can tell, no one has published a count of church websites like this: three complete denominational rosters, with every listed site followed. The method is below for anyone who wants to repeat it. Please tell us about any other research on church websites that you know about.

Set aside every broken link and look only at the 2,865 sites that answered a second time. The basics are good, and mostly due to the do-it-yourself website builders that host many of these sites. Nearly all read well on a phone and are served securely. Everything a visitor can check is in order, but everything that decides whether they turn up is not.

What the working sites are missing 

Start with the plainest case. Forty-two of these sites tell search engines not to list them at all. The instruction is a single word: noindexA single instruction in a page’s code, or in the reply from its server, telling search engines not to list it. Nothing about the page looks any different, and the page itself goes on working perfectly.Read the full entry →. Every builder has a switch that writes it, meant for the weeks before launch, and nothing on the finished page shows whether anyone switched it off. One setting in the wrong position keeps a church out of every search.

Nearly half of these sites carry no structured dataA small block of code in a page that states what you are, where you are and when you are open, in a form search engines read without guessing. Those facts can then appear in search results themselves, not just a link.Read the full entry →, the small block of markup that lets a search engine show a church's service times and address directly in results instead of just a link. The vocabulary is not Google's private invention: it is maintained jointly by Google, Bing, Yandex and Yahoo, and it contains a type for a church, which inherits a field for the hours it is open. Four search engines agreed a church should be able to publish when it meets. Google's own guidance is written in terms of eligibility — without the required properties, a page is not eligible for the enhanced result. It is missing from 1,368 church websites.

A third of homepages carry fewer than two hundred words. What kind of church is this? What is a Sunday morning like? Would I be out of place? Two hundred words is not enough to answer any of that, and it gives a search almost nothing to match against whatever someone typed.

The study cannot say why homepage content is often so thin. In our own work with churches, the two explanations we observe are that nobody was sure what to write, and that the page was designed to look good before anyone decided what it needed to say.

Two measures decide what a blind visitor gets. Almost a third of these sites have no main heading at all, and only 40% keep a clean heading structure, which is the map a screen reader uses to move through a page. Four sites in five leave at least one image undescribed. Most carry an empty alt attribute, which is the markup for a picture that does not matter. On a page of photographs from a Sunday, they do matter.

348 working sites show a copyright year of 2024 or earlier. 130 read 2020 or before, and 28 are frozen in 2020, the year everything stopped. The oldest reads 2003. Every one of those sites was answering requests when we checked.

A website can outlive the congregation that made it. Lifeway Research counts about four thousand Protestant churches closing in the United States each year — 4,500 in 2019, 4,000 in 2024, against roughly as many opening. No one takes the website down. The domain renews on a credit card someone set up years ago, so the host keeps answering, and the footer holds the year that someone last touched it. For some share of these 348, the date is accurate. It is the same reason directories keep listing churches that are not there.

Nearly a quarter of these sites arrive as a bare link when a member shares them. Social preview tagsA few lines of code in a page naming the title, description and image to use when somebody shares your link. Most apps read them to turn a pasted address into a preview card with a headline and a picture.Read the full entry → are what most apps read to decide what a link becomes: a title, a sentence and a photograph, or an address on its own. 77% of these sites carry them. The rest look plainest at the very moment somebody is recommending them.

About one site in four declares no canonical addressThe address you name as a page’s real home, stated in the page’s own code, for when the same content can be reached at several addresses. It asks a search engine to treat that one as the original.Read the full entry →, the line naming which version of a page is the real one when several answer. When the canonical address is missing, the search engine picks for itself. It might pick the domain you used before a rename, or the default your website builder handed you. Neither is what is printed on your sign.

Three sites in ten load at least one image, font or script over an insecure connectionWhen a secure (HTTPS) page loads an image, font, or script over an insecure (HTTP) connection. Browsers block the risky kinds outright and stop treating the rest of the page as fully secure.Read the full entry →. Only one stray file is enough for a browser to stop treating the page as fully secure. Most browsers now block the file rather than fetch it in the clear, so what a visitor sees is a photograph that never appears. The certificate is valid and the server is configured correctly, but the page is downgraded anyway.

The platform decides 

The general-purpose tools have won. Squarespace, WordPress and Wix account for more church websites between them than every church-specific platform combined, several times over. The market built for churches is a minority of the churches.

We expected the shortfalls above to be spread evenly, because they look like the kind of thing individual churches get wrong. On the single measure that most determines whether a church's service times can appear in results at all, the range between builders runs almost the whole scale.

The uncomfortable part is which platforms sit at which end. The general-purpose website builders do this automatically. The platforms built specifically for churches mostly do not. Squarespace and Wix publish structured data on every site we measured, without the customer ever knowing the word. Subsplash and Tithely — church platforms, sold to churches, for church purposes — publish it on three percent and two percent respectively.

This is not incapacity. Subsplash sets a search description on every site it hosts and a sitemap on 97% of them: the automated work is being done, and done well. What it does not emit is the one block of markup naming the organization as a church, giving an address, and stating when it meets — the three facts its customers exist to publish. The capability is demonstrated on every other measure.

Three thousand churches cannot each be asked to fix this. A handful of vendors could fix most of these problems for every church they host at once, and nobody would have to be told.

That chart counts whether a site emits structured data. It does not ask what the markup says, so we opened all 1,496 blocks and read them.

Presence is not usefulness. Of the 1,496 sites publishing anything, 83 identify the organization as a church — 3% of every working site. Nearly all name something, and one in ten of those is not the congregation: an acronym, a bare domain name, in one case the file name of a photograph. 716 carry an address. Only 202 break the address into the four fields a machine needs to drop a pin on a map.

And the platforms fail in opposite directions. Squarespace describes a business, with opening hours, and never says what kind of business: three of its 673 sites name a church, though 68% carry an address. Church.co names a church on 98% of its sites and gives an address on 2%.

Eighteen sites in the entire study publish all three of the things a stranger needs: that this is a church, where it is, and when it meets.

None of this is a character test on the churches. Nobody handed them the list above, and no congregation should be expected to know that search engines want specific fields according to shared standards. The platforms are a different matter. They employ people who work with web technologies for a living and they sell to churches specifically.

That leaves the part a church can act on without waiting for anyone. Most of what this section measured is in the checklist at the end of this study, divided between what a church can check for itself and what belongs to whoever maintains the site.

Run this twice a year, whether you serve a church or a nonprofit. Every item comes from something we found.

That is what can honestly be asked of a person with other work. It is also the wrong interval for the problem: a domain can lapse the week after you check, and nothing announces it.

No one writes it down. Then someone leaves. A single set of keys, in a single pair of hands, and a day that came when those hands moved on.

The language figures in Finding Four come from reading congregation names, and names find the better-resourced churches. One denomination sorts some congregations into language-based regional bodies, which let us check: among those 207, the ones names identify have no website 28% of the time; the ones names miss, 46%. Names report a gap three and a half times the rest of the denomination where the structure shows five and a half.

We wrote a classifier that reads real web page content instead of the names. That method puts the number of language-minority congregations at 566, with a range of 219 to 1,350. The classifier cannot see a church with no page to read, and those are the churches this finding is about.

Congregations are also frequently not one language but two or three, threading between a Korean service and an English one, Spanish preaching and English children’s ministry. A single language label would hide the true nature of the most interesting churches.