Research · Data study
2,000 sites · 50 states · July 2026
We audited 2,000 church websites across all 50 states, drawn from a roughly 27,000-church national list — every site fetched by an identified, robots-respecting crawler in two passes, and 99.9% of reachable sites adjudicated by an LLM judge. Here is what we found, gaps and all, with the full methodology below.
Headline findings
Headline percentages are of the 1,734 reachable sites the LLM judge confirmed as small churches (plausibly under 100 in attendance, single campus). 195 large or multisite churches were excluded from the headline block, and 1 site was unjudged. Of the 2,000 sampled sites, 1,930 were reachable; 19 disallowed crawling and were excluded from every metric.
For comparison, across all 1,930 reachable sites — small, large, and multisite together — the same metrics read: 94.8% HTTPS, 94% mobile viewport, 67.3% working giving, 74% updated within the year, and 83.8% findable service times. Larger churches pull almost every number up.
The lead finding
We scored every judged site on a five-point visitor-readiness rubric — one point each for findable service times, a findable physical address, plan-a-visit or what-to-expect content, a clear next step for a newcomer, and current content. Across the 1,930 judged sites the mean score is 3.25 out of 5, and only 51.6% score a 4 or 5. Read the other way: 48.4% of church websites are missing at least two of the five things a first-time visitor needs.
| Readiness score | Sites (share of judged) |
|---|---|
| 0 of 5 | 155 · 8% |
| 1 of 5 | 95 · 4.9% |
| 2 of 5 | 326 · 16.9% |
| 3 of 5 | 358 · 18.5% |
| 4 of 5 | 538 · 27.9% |
| 5 of 5 | 458 · 23.7% |
| Readiness score | Sites | Share of judged sites |
|---|---|---|
| 0 of 5 | 155 | 8% |
| 1 of 5 | 95 | 4.9% |
| 2 of 5 | 326 | 16.9% |
| 3 of 5 | 358 | 18.5% |
| 4 of 5 | 538 | 27.9% |
| 5 of 5 | 458 | 23.7% |
| All judged sites | 1,930 | 100% |
The five points split cleanly into logistics and hospitality. The logistics points are mostly in place: service times are findable on 84.1% of judged sites and a physical address on 85.3%. The hospitality points are not: plan-a-visit content exists on 56.9% of sites, a clear next step for a newcomer on 56.9%, and current content — anything dated within about three months — on just 41.2%. Most churches publish a website once. Far fewer keep one alive.
| Rubric point (1 point each) | % of judged sites |
|---|---|
| Service times findable | 84.1% |
| Physical address findable | 85.3% |
| Plan-a-visit / what-to-expect content | 56.9% |
| Clear next step for a newcomer | 56.9% |
| Current content (within ~3 months) | 41.2% |
Sermon content tells the same story. Of the judged sites, 435 (22.5%) have sermon or message content current within about two months, 133 (6.9%) have sermons that have gone stale, 730 (37.8%) publish no sermon content at all, and 632 (32.7%) have sermon content with no dates to judge by. And only 50.6% of sites name their pastor anywhere in the crawled text. The automated contact checks agree with the judge’s logistics read: 77.4% of sites show a phone number, 62.5% an email address, and 85% a street address.
The two-question test
Two yes/no signals separate a working church website from a parked one: was it updated within the year, and does online giving actually work? Crossing them splits the 1,930 reachable sites into four groups — and only 55.1% clear both bars.
| Quadrant | Sites | Share of reachable |
|---|---|---|
| Updated and giving works | 1,063 | 55.1% |
| Updated only | 366 | 19% |
| Giving only | 235 | 12.2% |
| Neither | 266 | 13.8% |
This is the study’s honest summary. Small church websites are rarely broken — HTTPS and mobile are near-universal — but only 55.1% function as an organ of the church: kept current and able to receive a gift. 366sites are tended but can’t take a gift, 235 can take a gift nobody is tending, and 266 (13.8%) do neither. A church website in 2026 is rarely broken. It is much more often abandoned.
Online giving
Across all reachable sites, 72.1% have a giving link on the homepage, but only 67.3% (1,298 sites) haveone that resolves when followed — “working” means the link was found and returned a successful response, minus sites the LLM judge marked as not actually working. Among working giving links, 15.8% (205) point at a recognizable giving processor by destination host.
| Giving destination host | Sites | Known processor |
|---|---|---|
| onrealm.org | 23 | Yes |
| give.tithe.ly | 21 | Yes |
| pushpay.com | 20 | Yes |
| tithe.ly | 20 | Yes |
| www.givelify.com | 17 | Yes |
| www.paypal.com | 17 | Yes |
| app.easytithe.com | 14 | Yes |
| secure.myvanco.com | 14 | Yes |
| www.eservicepayments.com | 13 | Yes |
| giving.myamplify.io | 11 | Yes |
| subsplash.com | 8 | Yes |
| www.shelbygiving.com | 7 | No |
| app.sharefaith.com | 4 | No |
| secure.accessacs.com | 4 | Yes |
| app.aplos.com | 3 | No |
Same-domain caveat: processor matching runs against the destination host only. A church whose giving page lives on its own domain (a /give page that embeds a processor further down) is counted as working but not as known-processor, so the processor share understates real processor adoption.
The digital divide is real, and giving is where it shows most. The 195 churches the judge classed as large or multisite take gifts online at 93.8%; the 1,734 small churches, at 64.2% — a 29.6-point gap. Freshness shows the same split (85.6% vs 72.8%). Service times are the one metric where small churches keep pace (83.9% vs 83.6%).
| Judged size class | n | Working giving | Updated ≤ 1 yr | Service times |
|---|---|---|---|---|
| Small churches (judged) | 1,734 | 64.2% | 72.8% | 83.9% |
| Large or multisite (judged) | 195 | 93.8% | 85.6% | 83.6% |
Freshness
Freshness is a composite: a site counts as updated within the year if any dated evidence — dated strings on the page, a sitemap lastmod, a non-dynamic Last-Modified header — falls within 365 days of the reference date, or the copyright year reads 2026 or later. A bare “© 2025” is ambiguous, so those cases were resolved by the LLM judge.
| Bucket | Sites |
|---|---|
| Updated within the year | 1,429 |
| Copyright 2025 only (still ambiguous) | 7 |
| Dated evidence, all stale | 264 |
| No dated evidence at all | 230 |
Where the evidence came from (a site can fire more than one source, so these multi-count): sitemap lastmod 1,020 sites, dated strings 501, copyright year 2026+ 332, Last-Modified headers 101. Of the ambiguous copyright-2025 cases, the judge resolved 21 to updated and 2 to stale. Last-Modified headers within two days of the crawl were treated as dynamically generated and dropped (238 sites).
Platform league table
Cut the same two outcome metrics — updated within the year, and working giving — by fingerprinted platform, and the spread is wide. Church-specific platforms sit at the top; the cheapest general-purpose builders sit at the bottom.
| Platform (n) | Updated within the year |
|---|---|
| The Church Co (n=48) | 97.9% |
| Subsplash (n=155) | 95.5% |
| Weebly (n=53) | 92.5% |
| Squarespace (n=198) | 91.4% |
| Tithe.ly Sites (n=45) | 86.7% |
| Wix (n=245) | 77.1% |
| WordPress (self-hosted) (n=277) | 73.3% |
| Other fingerprinted (n=168) | 71.4% |
| GoDaddy Websites + Marketing (n=67) | 61.2% |
Freshness league only — the table below adds working giving and service times.
| Platform | n | Updated ≤ 1 yr | Working giving | Service times |
|---|---|---|---|---|
| The Church Co | 48 | 97.9% | 89.6% | 91.7% |
| Subsplash | 155 | 95.5% | 93.5% | 98.1% |
| Weebly | 53 | 92.5% | 64.2% | 86.8% |
| Squarespace | 198 | 91.4% | 77.3% | 92.4% |
| Tithe.ly Sites | 45 | 86.7% | 91.1% | 91.1% |
| Wix | 245 | 77.1% | 64.9% | 82.4% |
| WordPress (self-hosted) | 277 | 73.3% | 67.1% | 84.5% |
| Other fingerprinted | 168 | 71.4% | 79.2% | 93.5% |
| GoDaddy Websites + Marketing | 67 | 61.2% | 61.2% | 86.6% |
An honest read: this is correlation, not causation. Church-specific platforms (Subsplash, The Church Co, Tithe.ly Sites) bundle giving into the product, so their giving numbers are partly definitional — and the kind of church that pays for a church-specific platform is already the kind that tends its website. Squarespace’s freshness lead likely reflects who chooses it as much as what it does. The platform does not make the church diligent; but the numbers say diligent churches cluster on certain platforms.
Ratings vs websites
We joined each sampled church to its Google Maps star rating (99.2% of the sample is rated; the mean rating is 4.72). The counterintuitive result: churches with a perfect 5.0 rating have the weakest websites of any rating band.
| Maps rating band | n | Working giving | Updated ≤ 1 yr | Service times |
|---|---|---|---|---|
| Below 4.5 stars | 232 | 60.7% | 71% | 81.3% |
| 4.5 to 4.9 stars | 1,338 | 73.2% | 76.7% | 85.9% |
| Perfect 5.0 stars | 414 | 51.4% | 66.8% | 78.6% |
| Unrated | 16 | 68.8% | 81.3% | 81.3% |
Churches rated a perfect 5.0 underperform the 4.5–4.9 band on every website metric: working giving 51.4% vs 73.2%, freshness 66.8% vs 76.7%. Before reading that as “beloved churches neglect their websites,” note the likelier explanation: a perfect 5.0 usually means a handful of reviews, and a handful of reviews usually means a very small congregation — so the 5.0 band is substantially a small-church band, and small churches trail on every digital metric (see the size divide above). The rating isn’t causing the website gap; size is likely driving both.
The measurement gap
Only 34% of reachable sites carry any analytics tooling we could detect — Google Tag Manager on 486 sites, gtag.js on 448, classic Google Analytics on 172, a Facebook pixel on 63. The other 66% have no idea whether anyone visits, what they look for, or where they leave. Every conversation about “does our website work” at two thirds of churches is happening without data.
| SEO basic | % of reachable | Note |
|---|---|---|
| Page title present | 97.6% | The one thing platforms handle automatically. |
| Meta description present | 62.4% | Median length 112 characters where present. |
| Exactly one H1 heading | 39.9% | 640 sites have none; 520 have several. |
| Open Graph title | 68.2% | Controls how a shared link looks in a feed or group chat. |
| Open Graph image | 47.3% | Under half of shared links show an image at all. |
| Favicon | 81.9% | The small polish signal — about 1 in 5 sites skips it. |
The pattern repeats down the checklist: whatever the platform does automatically gets done (97.6% have a page title), and whatever requires a human decision doesn’t (39.9% have exactly one H1; 62.4% have a meta description). Accessibility care shows the same neglect by proxy: across the 1,806 sites with images, the median share of images carrying alt text is just 40%, and only 23.7% of image-bearing sites exceed 80% coverage. (Alt-text coverage is a proxy for accessibility attention, not a WCAG audit.)
Social & streaming
Churches show up socially even when the website is neglected: 80.9% of reachable sites link at least one social profile, led overwhelmingly by Facebook. Livestream signals — an embedded player or streaming provider — appear on 29.8% (575 sites), a durable artifact of 2020 that most churches kept.
| Platform | Sites (share of reachable) |
|---|---|
| 1,460 · 75.6% | |
| YouTube | 1,084 · 56.1% |
| 774 · 40.1% | |
| X (Twitter) | 274 · 14.2% |
| TikTok | 46 · 2.4% |
Put the two gaps side by side and the priority inversion is plain: 80.9% of churches maintain a presence on someone else’s platform, while 66% can’t see what happens on their own.
Platforms & spend
We could fingerprint a platform on 65.1% of reachable sites (1,256); the remaining 34.9% are custom-built or unknown. Across the 1,010 fingerprinted sites on platforms with a published entry price, the median visible spend is $16/month — small churches are not, on the whole, overpaying for hosting.
| Platform | Sites | % of reachable | Entry price |
|---|---|---|---|
| WordPress (self-hosted) | 277 | 14.4% | $15/mo |
| Wix | 245 | 12.7% | $17/mo |
| Squarespace | 198 | 10.3% | $16/mo |
| Subsplash | 155 | 8% | Not published |
| GoDaddy Websites + Marketing | 67 | 3.5% | $12/mo |
| Weebly | 53 | 2.7% | $12/mo |
| The Church Co | 48 | 2.5% | Not published |
| Tithe.ly Sites | 45 | 2.3% | $19/mo |
| Duda | 41 | 2.1% | $25/mo |
| Clover Sites | 37 | 1.9% | $28/mo |
| FaithConnector | 27 | 1.4% | Not published |
| WordPress.com (hosted) | 24 | 1.2% | $8/mo |
| Webflow | 22 | 1.1% | $23/mo |
| Ekklesia 360 (MonkCMS) | 14 | 0.7% | Not published |
| E-zekiel (Salem Web Network) | 2 | 0.1% | Not published |
| Sharefaith | 1 | 0.1% | $69/mo |
Fingerprint limitations, disclosed: platforms are detected from asset hostnames and the generator meta tag, so self-hosted WordPress without a generator tag is undercounted, and Subsplash/Tithe.ly counts may be inflated where a custom site merely embeds their giving or media widget. Entry prices are each vendor’s published starting price, not what any specific church pays.
Does spending more buy a better website? A little, and mostly at the top: sites on platforms priced at $12/month or less have working giving at 62.5%, the $13–20 band at 70.5%, and the over-$20 band at 82.2%. Freshness climbs the same stairs (72.9% → 80% → 88.1%). The same caution as the platform table applies: churches that pay more are also churches that care more.
| Visible spend band | n | Working giving | Updated ≤ 1 yr |
|---|---|---|---|
| $12/mo or less | 144 | 62.5% | 72.9% |
| $13 to $20/mo | 765 | 70.5% | 80% |
| Over $20/mo | 101 | 82.2% | 88.1% |
By denomination & by metro
Denomination first, with a caveat that matters: the source list carries only two raw categories (“baptist church” and “christian church”), so our denomination field is a keyword normalization over that category plus the church’s name — churches with generic names land in “other,” and non-Baptist denominations are undercounted. Treat this cut as the roughest in the study.
| Denomination category | n | Working giving | Updated ≤ 1 yr | Service times |
|---|---|---|---|---|
| Baptist | 1,272 | 63.3% | 71.4% | 82.9% |
| Church of Christ | 38 | 40.5% | 62.2% | 86.5% |
| Other / generic name | 679 | 75.9% | 79.7% | 85.4% |
4 further categories covering 11sampled churches are withheld: at n < 6 a row stops being an aggregate and starts being a description of individual churches, which our naming policy rules out.
And by metro. The sample deliberately over-weights 29 metros (up to 30 sites each), which makes city-level cuts possible but small. Treat every row as directional, not statistically significant — a single church flips several points at these sample sizes.
| Metro | n | Working giving | Updated ≤ 1 yr | Service times |
|---|---|---|---|---|
| Atlanta, GA | 24 | 95.8% | 75% | 83.3% |
| Nashville, TN | 30 | 80% | 63.3% | 83.3% |
| Dallas, TX | 28 | 84% | 64% | 68% |
| Houston, TX | 30 | 82.8% | 69% | 79.3% |
| Austin, TX | 30 | 69.2% | 69.2% | 88.5% |
| Phoenix, AZ | 30 | 80% | 86.7% | 86.7% |
| Charlotte, NC | 28 | 74.1% | 85.2% | 81.5% |
| Raleigh, NC | 30 | 79.3% | 82.8% | 79.3% |
| Indianapolis, IN | 25 | 76% | 84% | 76% |
| Columbus, OH | 8 | 100% | 71.4% | 100% |
| Cincinnati, OH | 24 | 73.9% | 52.2% | 91.3% |
| Kansas City, MO | 12 | 58.3% | 66.7% | 83.3% |
| Minneapolis, MN | 30 | 70% | 66.7% | 76.7% |
| Tampa, FL | 30 | 65.5% | 62.1% | 79.3% |
| Orlando, FL | 23 | 55% | 65% | 85% |
| Jacksonville, FL | 30 | 81.5% | 74.1% | 81.5% |
| Birmingham, AL | 15 | 73.3% | 80% | 66.7% |
| Memphis, TN | 30 | 78.6% | 64.3% | 78.6% |
| Tulsa, OK | 30 | 69% | 75.9% | 75.9% |
| Oklahoma City, OK | 30 | 90% | 86.7% | 90% |
| Louisville, KY | 30 | 77.8% | 77.8% | 88.9% |
| Greenville, SC | 30 | 66.7% | 70% | 93.3% |
| Knoxville, TN | 27 | 70.4% | 70.4% | 81.5% |
| Chattanooga, TN | 30 | 73.3% | 80% | 80% |
| Richmond, VA | 25 | 73.9% | 78.3% | 82.6% |
| Pittsburgh, PA | 6 | 33.3% | 33.3% | 66.7% |
| Denver, CO | 15 | 93.3% | 86.7% | 93.3% |
| Los Angeles, CA | 15 | 80% | 73.3% | 86.7% |
| Seattle, WA | 8 | 87.5% | 87.5% | 100% |
n is sampled sites per metro; percentages are of that metro’s reachable sites. Each metro links to our notes on that city’s church web scene.
Citation terms
Every figure in this study is free to cite — in articles, podcasts, sermons, decks, academic work, or AI-generated answers — with attribution to Church Digital and a link back to this page. No permission or email is required. Please don’t present the numbers as your own measurement, and keep the denominators with the percentages where you can — the methodology above is part of the claim. Every published figure is an aggregate: no individual church is named, scored, or shamed, and citations should keep it that way.
Want ready-to-paste sentences? Every stat, quotable — /research/stats
Methodology
Everything below is what we would want to know before trusting someone else’s numbers.
Citing the study? Every stat, quotable — /research/stats
The frame is a roughly 27,000-church national list sourced from Google Maps (a lead-generation CSV), normalized, deduped by registrable domain, with obvious non-small-church rows excluded. From it we drew a deterministic stratified sample of 2,000 sites (seed 20260716, so the draw is exactly reproducible): up to 30 sites in each of 29 target metros, with a national fill capped at roughly 100 sites per state. All 2,000 were crawled on July 16, 2026 (single day, two passes: v1 15:10–16:17 UTC, v2 18:06–18:52 UTC); 1,930were reachable in the first pass. The study’s original 500-site draw is a strict subset of this sample.
The crawler identified itself honestly as ChurchDigitalResearch/1.0 (+https://churchdigital.com/research; logan@churchdigital.com), respected robots.txt (19 disallowing sites were recorded as excluded and never fetched), waited 1.5 seconds between requests to the same domain, timed out at 10 seconds, and requested at most ~6 (homepage, giving link, up to 4 subpages, sitemap). Two consecutive failures marked a site unreachable rather than retrying harder. Both crawl passes — v1 (core signals) and v2 (page-quality signals) — ran under exactly this etiquette, and the aggregate-only naming policy below is unchanged from the first edition.
A second crawl pass over the same 2,000 sites captured page-quality signals the first pass did not: page title, meta description presence and length, H1 count, Open Graph title and image, favicon, image count and alt-text coverage, social profile links by platform, livestream embed signals, phone / email / street-address presence, and analytics tooling (detected from script hosts and inline calls — Google Tag Manager, gtag.js, classic Google Analytics, Facebook Pixel, Plausible, Fathom). The v2 pass found 1,931 sites reachable — one more than v1, reachability being slightly time-of-day dependent — and v2 percentages use the v2 denominator.
Automated checks misread real websites in predictable ways, so an LLM judge (claude-haiku-4-5, temperature 0) reviewed 1,929 reachable sites (99.9% of reachable — above our 90% publication threshold) and corrected three signals where it disagreed: whether service times are actually findable (the regex proxy and the judge agree on 96.4% of judged sites), whether the giving path actually works, and how to read ambiguous copyright-2025-only freshness evidence (21 resolved to updated, 2 to stale). The judge also classified congregation size, which is what restricts the headline block to the 1,734 confirmed small churches.
A second judge pass (claude-haiku-4-5, temperature 0) scored visitor readiness over the v2 crawl’s stripped page text (homepage plus up to four subpages, capped at 12,000 characters). The rubric, verbatim from the judging prompt — each boolean worth exactly one point:
The judge was instructed to be conservative — “prefer false / ‘unclear’ over guessing” — the score is recomputed from the five booleans defensively, and the prompt forbids quoting any text from the sites. Both judge passes ran once with append-mode resume (no site judged twice), and each run prints a cost estimate computed from actual token usage; the study’s marginal cost is a function of pages crawled and tokens judged, not of any per-church decision.
Denomination and Google Maps rating come from the original source CSV, joined by registrable domain (2,000 of 2,000matched). Two caveats: the source carries only two raw categories (“baptist church” / “christian church”), so denomination is a keyword normalization over that category plus the church name and is the roughest field in the study; and no phone numbers, place URLs, or outreach state were carried over — only the category, the star rating, and the domain.
Nothing here is a rendering test, a full accessibility audit, a Core Web Vitals measurement, or a design judgment. The mobile figure is a viewport-tag proxy; the alt-text figure is an attention proxy, not a WCAG result. We did not test whether giving flows complete a transaction, only whether the path resolves. We did not measure traffic, search rankings, or whether anyone visits these sites at all — which the measurement-gap section suggests most churches don’t know either.
No individual church is named, ranked, or shamed — here or in anything we share on request. Every published figure is an aggregate over at least a metro-sized cut; denomination rows below n = 6 are withheld for the same reason. The point of the study is the state of the field, not a scoreboard.
Questions about the study
If you run one of these sites
We record a free five-minute video review of any church website — the giving path, the mobile flow, whether service times are findable, the same visitor-readiness rubric as the study, applied to your site. No pitch, no call required.