My Sunday Scorecard

Every claim, checked. Every sermon, scored twice.

Paste a public sermon link and get every claim traced to the text in its original language, scored against the church's own beliefs and the whole church's, and screened for high-control patterns.

Free to try. Results in a few minutes. Already scored? You'll see the saved scorecard.

Public sermons only. By submitting, you confirm the church or minister posted this video publicly on YouTube.

Two standardsOriginal-language checkClips, not hearsayRight of reply

The Church of Eleven22 · Joby Martin

The Arrest of Jesus, Matthew 26:47–56

Scored

Own doctrine

of 10

Whole church

14-seat panel

Control index

Watch

8:41

Jesus calling Judas "friend" is a last offer of forgiveness.

Stretched

18:14

"Put your sword back" read as a duty to pick it up. "Email me. I won't read it."

Misread

40:55

Christ returns with a tattoo on his thigh.

Misread

18 claims checked · 3 shown · Real scorecard from the Register

What you get

A scorecard you can check yourself

Every claim, timestamped

Each claim about Scripture, history or the faith, with the moment it was said and the clip to watch, right on the page.

The original language

Hebrew and Greek word studies, translations side by side, and how the church has read the passage through history.

Two standards

Scored first by the church's own public statements of faith, then against the confessions, literature and scholarship of nearly every Christian tradition.

The control screen

Ten weighted markers of high-control environments, from loyalty pressure to insulation from accountability.

Quote check

Every quote attributed to someone is checked against its real source. Misattributions are tracked by church and preacher.

Their best defense

For every finding we state the strongest case for the pastor before we answer it. No straw men.

The Register

Latest scorecards

Every scored sermon is filed by church and preacher, so patterns show across sermons. Doctrine scores are out of 10; the control index runs 0 to 100, lower is better.

Church · preacher	Sermon	Own doctrine	Whole church	Control
The Church of Eleven22 Joby Martin · Jacksonville, FL	The Arrest of Jesus 58:42 · 18 claims	7.5	6.0	30 Watch
Gospel in Life Timothy Keller · Redeemer Presbyterian	How to Find the Way 21:25 · 14 claims	9.0	7.0	4 Low
Word on Fire Bishop Robert Barron	A Living Sacrifice 14:12 · 14 claims	9.0	8.0	4 Low

Scoring this week

Shepherd's Watch

Where the control index runs high

A ranking of the churches whose sermons show the most high-control and manipulation patterns, measured across every reviewed sermon we've scored from them. A pattern over time, never one bad Sunday, and a screen of language, not a verdict on anyone.

Low0–15Watch16–35Elevated36–60High61–100

First rankings publish at 40 churches. 3 scored so far. Rankings need enough churches, and enough sermons from each, to be fair.

What we listen forWt

Insulation from accountability15

Pastor-centered preaching12

Loyalty pressure12

Brand over Christ10

Fear and us-versus-them10

Financial pressure10

Shame and the private life10

Staff used as object lessons8

Spiritual bypassing8

Delivery5

How it works

The same method for every pulpit

Transcribe and trace
We pull the full transcript and find every claim the sermon makes about Scripture, history or the faith.
Check the text
Each claim is checked in Hebrew or Greek, across translations, and against how the church has read it.
Score it twice
By the church's own public statements of faith, and by the confessions, literature and scholarship of nearly every denominational tradition across the whole church.
Screen for control
We weigh the markers of high-control environments and show the exact moments on video.

"Not many of you should become teachers, my brothers, because you know that we who teach will be judged more strictly."James 3:1

The show

Watch the breakdowns

Each week we take a real sermon, play the claims, check them on screen, and score it. Episodes play right here on the site.

For pastors

My Sermon Scorecard

Run your message before Sunday. See the critical issues, the claims that won't hold, and suggested rewrites while there's still time to change them. Private to you, never published.

Already scored? Pastors can claim their church and reply on the record. Replies appear next to the scorecard.

Claim your church

Your account

Follow your church

One account works here, on the CAN site and in the Concordance.

Get every new scorecard from churches you follow
Submit any publicly posted sermon by its YouTube link
Save scorecards and share them

Questions, answered

What Sunday Scorecard is, and what it isn't

What is Sunday Scorecard?

Commentary and analysis of sermons that churches and ministers have posted publicly, published by the Church Accountability Network (CAN). Each scorecard traces what was preached back to the cited texts, the church's own statement of faith and the wider Christian tradition, then gives our assessment with every source shown.

Scores, labels and rankings are our opinions. They rest on facts anyone can check: the speaker's own words, the moment each was said, and the texts and sources we compare them against.

How is a sermon scored?

We transcribe the public sermon, trace every claim about Scripture, history or the faith, check each one in the original languages and across translations, then score it twice: by the church's own public statements of faith, and by the confessions, literature and scholarship of nearly every denominational tradition across the whole church. The full method, weights and prompts are published under Transparency in Assessment.

Whose sermons can be scored?

Only sermons the church or minister has published on YouTube for anyone to watch. We do not score private, unlisted or members-only videos, recordings someone made from the pew, leaked material, or private messages and conversations. Submissions need a public YouTube link.

We discuss the public teaching of the person preaching. We do not score congregants, children, staff who appear on camera, or anyone's private life.

Is a score a judgment of the pastor?

No. A score measures one public sermon against the texts it cites and the standards it claims. It says nothing about the preacher's character, sincerity, salvation or ministry as a whole.

What is the control index? Does a high score mean abuse?

The control index is a 0 to 100 screen of the language in a public sermon for patterns that researchers and reporters on high-control groups have linked to risk, such as insulation from accountability, loyalty pressure, financial pressure and shame. Lower is better, and every flag links to the exact moment in the video.

It is not a finding, or an allegation, that abuse, misconduct or any crime has occurred, and we do not make such claims. One sermon is weak evidence, which is why Shepherd's Watch only looks at patterns across several reviewed sermons.

Who decides the score: people or AI?

AI drafts every scorecard, and a human editor reviews them, some before publication and some after. Every scorecard says which. AI models (Claude, by Anthropic) pull out the sermon's main claims, run the text studies, check quotes and voice a panel of seats, each representing a Christian tradition or academic discipline and working from its published sources. The seats are AI personas: they are not real people and do not speak for any church, scholar or institution.

Can it get things wrong?

Yes. AI can misread a passage, mishear a transcript or overstate a pattern. Every quote is matched word for word against the transcript, every claim links to the moment it was said, times marked ~ are approximate, and anything we cannot source is labeled unsourced, not false. If we got something wrong, tell us and we will fix it.

I'm the pastor or the church. How do I respond?

Use the right of reply. You can claim your church and reply on the record; your response appears next to the scorecard and on any Shepherd's Watch row that includes it. When a reply shows a factual error, we correct the scorecard and add a dated note saying what changed. Disagreement is published too. Pastors can also score their own message privately before Sunday.

What if a video was posted by mistake, or has been taken down?

Tell us. If a church did not mean a sermon to be public, or has removed it, we take down the scorecard and any clips from it. Clips play through YouTube's own player, so they stop playing once the video is made private or deleted. Clips in our produced episodes are removed on request.

How are video clips used?

On a scorecard, clips play in YouTube's embedded player at the moment being discussed; the video stays hosted by YouTube on the church's channel. Our produced episodes use short excerpts for commentary and criticism, always credited and linked to the full sermon so viewers can hear the context.

Why score a sermon twice?

Fairness. A Baptist sermon is first measured against Baptist confessions and the church's own “What we believe” page, a Catholic homily against the Catechism, and so on. The second score shows where the sermon sits in the wider Christian tradition. A sermon can be faithful to its own church and still be contested elsewhere, and the scorecard shows both.

How does a church end up on Shepherd's Watch?

Only through a pattern. A church needs at least three scored sermons in the last twelve months, the public list counts only reviewed scorecards, and rankings do not publish until 40 churches are scored. Churches are ranked by the Watch Index, whose weights are published below. Every listed church can reply, and we pin its reply.

Which sermons have been scored?

The Register lists every scored sermon by church and preacher, starting with The Church of Eleven22 (The Arrest of Jesus), Timothy Keller (How to Find the Way) and Bishop Robert Barron (A Living Sacrifice).

Is Sunday Scorecard connected to any church or denomination?

No. We are independent of every church, denomination and ministry we write about. Church and ministry names appear only to identify the sermon under discussion and do not imply any affiliation or endorsement.

The whole method, open source

If we ask churches to be transparent, we should be too. Everything that produces a score is published here: the weights, thresholds, panel, word lists and the exact instructions the AI models receive. It is generated from the same engine that scores every sermon. Anyone may read, test, critique and reuse it under CC BY 4.0, with credit to the Church Accountability Network.

01How a sermon is scoredTen steps, from public link to human review.

Public source only. The tool takes a public YouTube link, pulls the published transcript and the church's own "What we believe" page, and is designed to stop if the video is marked private or unlisted.
Claims. The 10 to 14 main claims are pulled out, each with the preacher's exact words and a timestamp, and scored 0 to 10 on four measures.
Text study. Cited passages are checked in Hebrew, Aramaic or Greek, against the oldest manuscripts, across translations and through church history.
Own standard. Every claim is judged against the church's own statement of faith and its tradition's confession.
Panel. 20 seats, one per major tradition and discipline, each give a verdict, a score and a stance on every claim.
Positions. The sermon is placed on 25 theological questions where traditions differ. A question the sermon did not raise is marked silent.
Control screen. 10 weighted patterns of high-control language, plus tone.
Quote check. Every attributed quotation is traced to its documented source.
AI check. Measured writing signals, and quotes or citations we could not find in any source, are weighed against signs of a human hand.
Human review. Reports may be published before an editor reviews them. Every report shows its review status, and only reviewed reports count toward the public Watch list. The church can reply at any time.

Every quote shown in a report must match the transcript word for word. Quotes that do not match are dropped, never paraphrased into quotation marks.

02The Sunday ScoreFive weighted measures, with a live self-test.

Self-test passed: these weights match the scoring code

35% Textual fidelity	The cited text supports the claim in its own context.
20% Support across the Bible	Independent witness across the whole Bible.
20% Historical consensus	10 = ecumenical creed, 5 = one major tradition, 1 = one teacher.
10% How widely agreed	10 minus how sharply the claim is disputed across traditions.
15% Panel mean	Average whole-sermon score from the twenty seats. When no panel runs, the other four are rescaled to 100%.

Bands, on a 0 to 100 scale: Sound 85+ · Mostly sound 70+ · Mixed 50+ · Strained 30+ · Unsupported 0+. The own-standard score (0 to 100) is reported beside it and is not blended in.

03Control and manipulation screen10 weighted patterns. A risk screen, not a verdict.

This is a risk screen of the words in one public sermon. It is not a finding that abuse or misconduct happened. Each pattern is scored 0 to 10, then weighted to a 0 to 100 index.

10 Brand over Christ	the church, its name, campuses, vision, growth metrics, the pastor's books/tour/podcast promoted or named more than Christ; 'this house' language; celebrity framing
12 Pastor-centered narrative	illustrations that make the pastor the hero, martyr or prophet; 'God told me' as authority for institutional decisions; performed humility that still centers the self; name-dropping
15 Insulation from accountability	'touch not the Lord's anointed'; critics labeled divisive, gossips, bitter, jealous, wounded, a Jezebel spirit, wolves; Matthew 18 used to forbid public truth-telling; media, bloggers, former members dismissed; elders described as deferring to vision
12 Loyalty, unity and covenant pressure	submission and honor culture; 'unity' invoked to end questions; leaving framed as rebellion, covenant-breaking or spiritual danger; 'buy-in', 'all-in', 'get on the bus'
10 Us-versus-them and fear	persecution narrative, enemies inside and outside, culture-war contempt, apocalyptic urgency used to drive compliance
10 Financial pressure	seed/sow/harvest, tithe as the test of faith, giving tied to blessing or curse, campaign urgency, guilt or flattery around giving, prosperity framing
10 Shame and private-life control	public shaming of named or identifiable sins; intrusive control of dating, marriage, sex, family, career; gendered control; 'accountability' meaning surveillance
8 Staff and volunteer culture as object lesson	stories about staff who 'didn't fit', were 'let go' or 'lacked loyalty'; burnout glorified as sacrifice; 'we don't talk about that' norms; volunteers as fuel
8 Minimizing abuse / spiritual bypassing	'forgive and move on'; 'handled internally'; victims asked to protect the church's witness; abuse reframed as 'conflict', 'a hard season' or 'attack of the enemy'
5 Delivery: contempt, mockery, intimidation, flattery	HOW it is said: sarcasm or mockery at opponents, yelling, grandiosity, contempt, flattering the room, demands for applause or response

Index bands: Low 0 to 15 · Watch 16 to 35 · Elevated 36 to 60 · High 61 to 100. Tone is read on warmth, contempt, fear, urgency, flattery, grievance, certainty, humor. When no audio or video analysis is supplied, tone is inferred from the words alone and held loosely.

Sources the pattern profile is drawn from: investigative reporting on church abuse (The Roys Report), Robert Lifton's criteria for thought reform, Steven Hassan's BITE model, Wade Mullen on impression management, Chuck DeGroat on narcissistic church systems, and Diane Langberg on spiritual abuse. The screen also records what is reassuring, such as a named plural eldership or an open invitation to ask questions.

04Shepherd's Watch IndexHow churches are ranked, using patterns only.

40% Median control index	The weighted control screen, 0 to 100.
25% Shortfall from a sound sermon	100 minus the median Sunday Score.
15% Serious red flags	Per sermon, capped at two.
10% Recurring patterns	Control patterns scoring 5+ in two or more sermons, capped at four.
10% Misread scripture	Share of claims whose cited text says something different in context.

A church needs at least 3 scored sermons in the last 12 months. Medians are used rather than averages, so one sermon cannot move a church far. The public list counts reviewed reports only. Every listed church can reply, and its reply is pinned.

05The panel20 AI seats, their sources and their calibration.

The seats are AI personas. Each is built from a tradition's published sources and does not speak for any real person or institution. Every seat is tested on the same 5 calibration claims, so you can see how it leans before it judges a sermon.

Roman Catholic theologianEastern Orthodox theologianAnglican theologianLutheran theologianReformed theologianBaptist theologianWesleyan-Methodist theologianClassical Pentecostal theologianAnabaptist theologianHistoric Black Church theologianNondenominational evangelical pastor-theologianGlobal South theologianProgressive mainline theologianHebrew Bible scholarNew Testament scholarChurch historianJewish scholar of the Hebrew BibleCritical historian of early ChristianityPastoral care and spiritual-abuse specialistSteelman advocate

Download every seat profile (in the method file), with its sources, assumptions and calibration answers.

Calibration claims: K1 Jeremiah 29:11 is God's personal promise that your plans will prosper. K2 Bring the full tithe and God will open the windows of heaven over your finances (Malachi 3:10). K3 Christ died for our sins according to the Scriptures (1 Corinthians 15:3). K4 'Touch not my anointed' (Psalm 105:15) means you must not question your pastor. K5 Baptism is an outward sign only; it does nothing.

06Own standard and baselinesEach church is judged first by its own confession.

Each church is measured first by what it says it believes. Its "What we believe" page is pulled when it can be found. When it cannot, the report says so and infers the tradition. Then the governing confession for that tradition is applied:

Southern Baptist → Baptist Faith & Message 2000
Other Baptist → 1689 London Baptist Confession / New Hampshire Confession / the church's own statement
PCA, OPC, EPC, ARP → Westminster Confession + Larger & Shorter Catechisms
PC(USA) → Book of Confessions (Westminster, Heidelberg, Barmen, Confession of 1991)
Reformed (CRC, RCA, URCNA) → Three Forms of Unity (Belgic, Heidelberg, Dort)
Roman Catholic, Latin rite → Catechism of the Catholic Church, Vatican II, Roman Missal; Eastern Catholic rites → same catechism plus the rite's own liturgical tradition
Eastern Orthodox → the Seven Ecumenical Councils, the Fathers, Confession of Dositheus, the Divine Liturgy
Oriental Orthodox → first three councils, Coptic/Armenian/Syriac tradition
LCMS, WELS → Book of Concord (unaltered Augsburg, Formula of Concord)
ELCA → Book of Concord, with its own social statements
Anglican/Episcopal → 39 Articles, Book of Common Prayer, Lambeth Quadrilateral; ACNA → Jerusalem Declaration
United Methodist → Articles of Religion, Confession of Faith, Wesley's Standard Sermons & Notes; Free Methodist / Wesleyan / Nazarene → their Articles of Faith (Nazarene Manual)
Assemblies of God → 16 Fundamental Truths
Church of God (Cleveland) → Declaration of Faith
Foursquare → Declaration of Faith
COGIC → Statement of Faith
Pentecostal Holiness → Articles of Faith
Vineyard → Statement of Faith, Core Values
Calvary Chapel → Calvary Chapel Distinctives
Acts 29 / Gospel Coalition churches → TGC Confessional Statement
Non-denominational evangelical → the church's own statement, read against the NAE Statement of Faith and the Lausanne Covenant
Hillsong / Bethel / Elevation-style → the church's own statement, read against the Apostles' & Nicene Creeds
Word of Faith (Rhema, Kenneth Copeland Ministries) → their own statements
Mennonite / Anabaptist → Confession of Faith in a Mennonite Perspective (1995), Schleitheim
Churches of Christ / Christian Church → "no creed but Christ": judge by their stated NT-restorationist commitments
Seventh-day Adventist → 28 Fundamental Beliefs
Evangelical Free → EFCA Statement of Faith 2008
Christian & Missionary Alliance → Statement of Faith (Fourfold Gospel)
Salvation Army → 11 Doctrines
Quaker → Richmond Declaration (evangelical) or none (liberal)
Latter-day Saints → Articles of Faith, Standard Works (note: outside Nicene Christianity)
Jehovah's Witnesses → Watchtower doctrine (note: outside Nicene Christianity)
Unitarian Universalist → Seven Principles (note: non-creedal, non-Nicene)

Custom runs can measure against a chosen baseline:

text	Score 'textual fidelity' strictly against what the cited passage says in its literary and historical context, in the original language where it matters. Do not credit a reading because a tradition holds it.
nicene	Score fidelity against the ecumenical consensus of Nicaea and Chalcedon plus the plain sense of the text; treat later denominational distinctives as contested rather than wrong.
reformed	Score fidelity against the Reformed confessions (Westminster Standards, Canons of Dort) and their exegesis.
wesleyan	Score fidelity against Wesleyan-Arminian theology (Wesley's sermons, Articles of Religion).
catholic	Score fidelity against Catholic magisterial teaching (Catechism, councils) and the Church Fathers.
orthodox	Score fidelity against Eastern Orthodox tradition (the seven councils, the Fathers, the liturgy).
pentecostal	Score fidelity against classical Pentecostal theology (Assemblies of God / Church of God statements), distinguishing it from Word of Faith teaching.
jewish	Score fidelity against a Second Temple / rabbinic Jewish reading of the Hebrew Bible; flag where Christian readings exceed what the Hebrew text carries.

07Quote and AI checksStatus labels, word lists and their limits.

Quote check. Each attributed quote is labeled Accurate, Misattributed, Not in the Bible, No known source, Altered wording, Fair paraphrase. "Misattributed" or "Not in the Bible" is used only when the documented record supports it. Otherwise the quote is labeled unsourced.

AI check. AI detection is unreliable, especially on spoken transcripts. The result is a signal, never proof, and it is phrased as a gracious question rather than an accusation. Measured: AI-typical phrases per 1,000 words, spontaneity markers per 1,000 words, and sentence-length variation, compared with the same church's other sermons where available. Levels: Low (Little in the text suggests AI writing.) Uncertain (Mixed or weak signals. Not enough to say either way.) Elevated (Several signals line up. Worth a gracious question.) Strong (Many signals, including fabrications. Still not proof.)

Phrases counted:

delvedelvestapestrytestament tobeaconrealmembarknavigate thenavigating thein today's fast-pacedfast-paced worldin a world whereit's not justnot just aboutat its coreprofoundresonateresonatesfosterunlockunleashjourney of faithlet's divedive deepdive intohere's the thingin conclusionultimately,moreoverfurthermoreadditionallyin essencea reminder thatserves as a reminderstands asplays a crucial rolecrucialpivotalmultifacetedintricatenuancedembrace thetransformativevibrantever-evolvinglandscape of

Signs of a human hand:

umuhyou knowamencome ony'allturn to your neighborcan i getsomebodylistenhold oni'm sorrywaitlook atraise your handgive it uplaughlaughter[laughter][applause]

08The exact promptsWord for word, as the models receive them.

These are the exact instructions the models receive for every sermon. Bracketed items are filled in with the sermon being scored.

Download all 8 prompts (method.json)

09Limits and changesWhat the method cannot do, and every version.

Analysis is drafted by large language models (Claude, by Anthropic). Models can be wrong, and the safeguards above are there to catch it.
Transcripts come from YouTube captions and can mishear words. Timestamps marked ~ are approximate.
A single sermon is a small sample. Patterns across sermons carry more weight than any one score.
The panel balances traditions on purpose, so a sermon that is faithful to one tradition can still score as contested.
Theological disagreement is not error. The report separates what the text says from what a tradition holds.

Changes to the method

v1.1 · 2026-10-07	Method published under CC BY 4.0. Public-video-only intake (attestation, private/unlisted block, link required). Opinion and commentary notice on every report. Public Shepherd's Watch counts reviewed reports only. Method published in full.
v1.0 · 2026-10-07	First release: two standards, twenty-seat panel, control screen, quote and AI checks, Watch Index.

Opinion and commentary

Sunday Scorecard publishes analysis, opinion and commentary on sermons their speakers chose to make public. Ratings, labels and rankings are our opinions, based on the facts disclosed in each scorecard.

No allegations of misconduct

We do not allege that any person has committed abuse, a crime or misconduct. Control and AI checks measure language and writing patterns in a public sermon only.

Public content only

We only analyze videos the church or minister has posted publicly. By submitting a link you confirm that it is. We may decline any submission and remove any scorecard.

Fair use

Short quotations and embedded clips are used for criticism, commentary, teaching and news reporting, in line with fair use (17 U.S.C. § 107). Videos remain hosted by YouTube on the owner's channel.

Names and marks

Church, ministry and denomination names are used only to identify the sermon under discussion. We are independent; no affiliation or endorsement is implied.

AI-assisted analysis

Scorecards are drafted with AI models and checked against the transcript. Panel seats are AI personas, not real people. Each scorecard shows whether a human editor has reviewed it.

Corrections and takedowns

Corrections are made promptly with a dated note. Requests for a reply, correction or takedown: mysundayscorecard.com/reply.

Not advice

Nothing here is pastoral, legal, medical or psychological advice. If you or someone else is in danger, contact local emergency services.