📌 The Swipe File

political ad inspiration · sourced from Meta Ad Library 📊 Analytics ⚖ Compare 📄 Playbook 🔑 Access
← gallery · Analytics (whole library)
122,038
ads
17,999
unique creatives
of 122,038 ads
~$86,300,231
est. total spend
1%
currently active
15+
advertisers
5d
median lifespan
17,995
creatives AI-analyzed
🧠 Full image+text analysis coverage
47% · 12,879 of ~27,678 real creatives (photo + copy) fully analyzed
Text/message analysis covers more (17,995 unique wordings); the visual half is complete on 12,879 so far. ~4,385 were analyzed on blank screenshots (excluded from visual charts, being re-analyzed).

What actually works

Each creative's tags vs. how its ads performed. Typical creative runs 12 days and reaches 39,905 — “lift” is a tag's average vs. that baseline (creative-level, min 40 creatives).
🟢 Extends ad life (longevity = proven creative)
survey poll +116% · 25d · reach -13%
list building survey +92% · 23d · reach -12%
sign petition +75% · 21d · reach -21%
social security medicare +73% · 20d · reach +35%
housing +67% · 20d · reach +11%
jobs labor +51% · 18d · reach +86%
immigration +40% · 16d · reach +71%
ominous threatening +30% · 15d · reach +49%
🔴 Dies fast (under-performs on longevity)
image heavy -48% · 6d · reach -38%
gotv mobilization -45% · 6d · reach -38%
unknown -40% · 7d · reach -62%
vote -37% · 7d · reach +9%
no text -33% · 8d · reach -89%
unknown -33% · 8d · reach -89%
education -32% · 8d · reach -10%
volunteer -30% · 8d · reach -48%
By message type
valuecreativesdays livereach
list building survey 1,931 23d +92% 35,010 -12%
fundraising 7,441 11d -4% 30,663 -23%
persuasion 4,446 11d -8% 63,138 +58%
awareness 1,852 9d -23% 39,384 -1%
gotv mobilization 1,589 6d -45% 24,731 -38%
By visual framing
valuecreativesdays livereach
issue advocacy 1,369 14d +20% 32,865 -18%
attack opponent 2,435 14d +15% 40,026 +0%
contrast comparison 3,025 12d +3% 41,829 +5%
informational 518 12d +1% 19,186 -52%
positive candidate 9,912 11d -7% 41,343 +4%
By emotional appeal
valuecreativesdays livereach
empathy 1,481 13d +12% 54,353 +36%
anger 2,494 13d +11% 48,881 +22%
fear 2,155 13d +11% 41,024 +3%
pride 2,783 12d +1% 47,842 +20%
urgency 6,394 11d -7% 31,562 -21%
hope 1,408 11d -9% 36,635 -8%
nostalgia 159 9d -26% 24,221 -39%
neutral 385 9d -27% 19,520 -51%
By production tier
valuecreativesdays livereach
professional 1,742 15d +28% 76,909 +93%
semi pro 9,323 13d +13% 49,642 +24%
diy amateur 1,646 10d -19% 34,466 -14%
unknown 4,548 8d -30% 7,740 -81%
By visual format
valuecreativesdays livereach
news screenshot 664 15d +28% 67,615 +69%
photo 2,434 14d +21% 38,826 -3%
mixed 1,539 14d +19% 38,715 -3%
graphic text card 3,056 14d +18% 32,233 -19%
illustration 64 13d +12% 20,779 -48%
video frame 4,480 12d -2% 75,029 +88%
social screenshot 621 11d -4% 54,016 +35%
unknown 4,390 8d -33% 4,548 -89%
By call-to-action
valuecreativesdays livereach
survey poll 405 25d +116% 34,884 -13%
sign petition 1,513 21d +75% 31,620 -21%
learn more 2,707 13d +10% 77,620 +95%
donate 7,457 11d -4% 30,542 -23%
none 3,502 8d -28% 34,190 -14%
volunteer 173 8d -30% 20,770 -48%
vote 1,502 7d -37% 43,640 +9%

How to read this: outcomes are observational, not causal — a tag that correlates with longer runs isn't proven to cause them (incumbency, timing, and spend all confound). Longevity is right-censored (still-running ads haven't finished), and reach uses Ad-Library range midpoints. Treat as directional signal, strongest where longevity and reach move together (e.g. issue ads, ominous mood).

Dataset at a glance
2018-11-06
earliest ad
2026-06-30
latest ad
2026-07-02
last captured
~$86,300,231
ad spend represented
2,824M
impressions represented
234
candidates
287
advertiser pages
Volume vs. variety
Deduplication (distinct creatives behind the ads)
6.8× ads per unique creative
122,038 ads collapse to 17,999 creatives — campaigns mass-produce variants (esp. fundraising).
Media / format mix (4480 video)
video frame 4,480
unknown 4,400
graphic text card 3,059
photo 2,434
mixed 1,539
social screenshot 1,087
news screenshot 921
illustration 64
meme 11
Variant churn — most ad-variants per creative (who A/B-tests hardest)
candidateadscreativesads / creative
OR — Jeff Merkley 6,904 164 42.1×
SC — Annie Andrews 3,249 148 22.0×
IA — Josh Turek 4,008 201 19.9×
NM — Ben Ray Luján 1,178 65 18.1×
TX — John Cornyn 6,061 337 18.0×
WY — Cynthia Lummis 859 51 16.8×
MI — Abdul El-Sayed 6,355 394 16.1×
ID — Jim Risch 476 31 15.4×
SC — Lindsey Graham 8,528 554 15.4×
KY — Amy Mcgrath 5,899 418 14.1×
NH — Jeanne Shaheen 1,000 80 12.5×
CO — John Hickenlooper 10,956 935 11.7×

Message mix over time

By ad volume per month — watch fundraising cede ground to persuasion & GOTV as the cycle matures.
Message type by month fundraisingpersuasiongotv mobilizationawarenesslist building surveyother
2025-07 1,979
2025-08 2,688
2025-09 2,825
2025-10 2,869
2025-11 1,690
2025-12 2,136
2026-01 2,370
2026-02 3,248
2026-03 4,749
2026-04 4,029
2026-05 4,978
2026-06 4,735
Attack-ad rate by month (% framed as attack_opponent)
2025-07 10.8%
2025-08 9.0%
2025-09 8.7%
2025-10 6.6%
2025-11 9.1%
2025-12 7.7%
2026-01 13.8%
2026-02 16.6%
2026-03 11.6%
2026-04 7.6%
2026-05 14.6%
2026-06 11.2%
Issue salience by month (top issues — % of that month's ads)
070809101112010203040506
democracy elections 454244514856454250504551
candidate character bio 17191118175121616171819
other 58139101115121110129
economy taxes 116977851110987
healthcare 71110657423545
immigration 210111861111
New creatives first seen per month (corpus growth — the refresh case)
2025-07 297
2025-08 317
2025-09 456
2025-10 454
2025-11 350
2025-12 374
2026-01 444
2026-02 674
2026-03 719
2026-04 686
2026-05 799
2026-06 621
Library & performance
Message type(keyword classifier)
Fundraising 64,957
Persuasion / Other 34,272
List-building 16,164
GOTV / Vote 5,706
Survey 880
Volunteer 59
Platform mix
facebook 113,083
instagram 89,274
messenger 159
audience network 148
threads 6
whatsapp 2
Spend tiers
< $100 82,258
$100–1K 28,700
$1K–10K 9,665
$10K–50K 1,222
$50K+ 193
Audience delivery (age · gender share %)
65+ · female2355796%
65+ · male1644468%
55-64 · female1085518%
55-64 · male914932%
45-54 · female607181%
45-54 · male586924%
35-44 · male498193%
35-44 · female488342%
25-34 · male420805%
25-34 · female386574%
18-24 · female169047%
18-24 · male161896%
Top delivery regions
Texas8887
California5720
Colorado4992
Michigan4363
Florida4118
North Carolina3826
Ohio3492
Kentucky3409
New York3217
Massachusetts3059
New Hampshire3059
Minnesota2632
Top advertisers by spend
advertiseradsest. spend
Jon Ossoff3545~$13,254,327
James Talarico856~$9,550,472
Lindsey Graham8528~$8,375,886
Amy McGrath5899~$6,248,900
Colin Allred4968~$5,052,166
Sherrod Brown4263~$4,701,418
John W. Hickenlooper10950~$3,639,375
Graham Platner543~$2,959,078
Josh Weil - Florida Progressive Democrat457~$2,562,871
John Cornyn6035~$2,185,082
Roy Cooper2876~$1,979,212
Dr. Annie Andrews3249~$1,752,725
Mallory McMorrow636~$1,739,882
Mary Peltola2018~$1,033,991
Susan Collins1303~$961,098
How long ads run — median 5d · mean 13 · longest 364d. A long-running ad is a proven ad.
Lifespan distribution
≤ 7 days 74,362
8–30 days 34,676
31–90 days 10,169
90+ days 2,831
Longer ads reach far more (avg per ad)
lifespanavg reachavg spend
≤ 7 days9,749$301
8–30 days39,146$1,129
31–90 days53,972$1,882
90+ days68,030$1,974
Longest-running ads (evergreens — click)
300dColin Allred20,000–24,999
299dColin Allred8,000–8,999
296dBluejay Rising250,000–299,999
294dColin Allred7,000–7,999
294dColin Allred7,000–7,999
294dColin Allred15,000–19,999
294dColin Allred9,000–9,999
285dLogan for Kentucky8,000–8,999
285dLogan for Kentucky5,000–5,999
285dLogan for Kentucky8,000–8,999
285dLogan for Kentucky4,000–4,999
285dLogan for Kentucky4,000–4,999
Ads created per month (campaign ramp)
6,634 0 1,979 2025-07 2,688 2,825 2025-09 2,869 1,690 2025-11 2,136 2,370 2026-01 3,248 4,749 2026-03 4,030 4,978 2026-05 4,735 2026-06

📝 Text & messaging analysis

What each ad says — issue, framing, ask. From the copy (read in context of the image). 17,995 unique creatives analyzed.
Issue / topic
democracy elections 6,748
candidate character bio 2,577
other 2,082
economy taxes 1,186
healthcare 859
immigration 508
jobs labor 494
energy environment 471
abortion 415
foreign policy defense 405
crime safety 386
education 277
agriculture rural 227
guns 223
Issue by est. spend($ behind each topic)
democracy elections 45,444,365
other 9,867,709
economy taxes 9,096,645
candidate character bio 9,087,030
healthcare 3,340,841
jobs labor 2,248,107
immigration 1,473,833
abortion 1,142,684
foreign policy defense 979,107
energy environment 755,283
crime safety 742,828
social security medicare 706,848
Framingpositive / contrast / attack
positive candidate 7,456
contrast comparison 2,221
attack opponent 1,896
issue advocacy 1,061
informational 245
Message type
fundraising 7,441
persuasion 4,445
list building survey 1,931
awareness 1,852
gotv mobilization 1,589
Emotional appeal
urgency 6,394
pride 2,783
anger 2,494
fear 2,155
empathy 1,480
hope 1,408
neutral 385
nostalgia 159
Call to action
donate 4,675
none 2,618
learn more 2,564
sign petition 1,319
vote 1,227
survey poll 323
volunteer 153
Top attack targets(who ads name/criticize)
Donald Trump 602
Mitch McConnell 318
Ted Cruz 248
Susan Collins 232
Cory Gardner 165
Joe Biden 158
Democrats 145
Republicans 136
Lindsey Graham 110
Paul LePage 107
Mike Rogers 101
John Sununu 86
🔀 Issue × framing how each topic is pitched — darker = more creatives
attack
opponent		contrast
comparison		positive
candidate		issue
advocacy		informational
democracy elections 995 1831 3205 368 349
candidate character bio 171 178 2226 2
other 136 70 1688 81 107
economy taxes 220 250 573 109 34
healthcare 190 182 340 141 6
immigration 143 91 172 101 1
jobs labor 31 52 368 43
energy environment 66 64 213 121 7
🥊 Who goes negative share of a candidate's creatives that are attack-framed
FL — Ashley Moody (12/12) 100%
AR — Tom Cotton (122/238) 51%
KY — Nate Morris (5/10) 50%
SC — Mark Lynch (22/50) 44%
NH — Chuck Morse (33/76) 43%
MN — Adam Schwarze (3/7) 43%
TX — Jasmine Crockett (16/38) 42%
IL — S Krishnamoorthi (97/239) 41%
IA — Thomas Laehn (2/5) 40%
TN — Robert Jones (4/11) 36%
DE — Chris Coons (181/500) 36%
LA — Eric Skrmetta (6/17) 35%
RI — Allen Waters (56/171) 33%
KY — Amy Mcgrath (134/418) 32%
OK — Markwayne Mullin (17/54) 31%
📈 What's working avg lifespan (red) & reach (gold) of ads carrying each tag — longer/wider = better
By framing
valuecreativesavg days liveavg reach
contrast comparison 3081 11d 26,107
positive candidate 10211 11d 24,963
attack opponent 2589 13d 23,650
issue advocacy 1451 20d 16,909
informational 663 20d 14,508
By emotional appeal
valuecreativesavg days liveavg reach
pride 2803 13d 37,934
empathy 1511 17d 31,408
nostalgia 162 10d 28,915
anger 2635 15d 26,425
fear 2190 11d 24,017
hope 1431 12d 21,984
urgency 6716 11d 20,161
neutral 547 22d 14,353
By issue
valuecreativesavg days liveavg reach
jobs labor 506 16d 54,421
veterans 158 16d 53,608
crime safety 397 10d 40,807
agriculture rural 232 20d 36,000
economy taxes 1258 10d 31,999
housing 69 16d 31,858
candidate character bio 2647 12d 30,049
education 306 10d 29,416
immigration 512 31d 28,080
social security medicare 183 17d 24,846
healthcare 930 12d 24,099
guns 224 11d 22,297
abortion 429 12d 21,537
democracy elections 7004 11d 20,530
energy environment 478 21d 19,293
other 2254 17d 17,284

🎨 Visual & creative analysis

How each ad looks — imagery, people, symbolism, craft. From a vision pass over the creative.
Format
video frame 4,479
graphic text card 3,056
photo 2,434
mixed 1,539
news screenshot 664
social screenshot 621
illustration 64
unknown 11
meme 11
Production tierpro vs DIY
semi pro 9,312
professional 1,741
diy amateur 1,624
unknown 202
Primary subject
candidate 6,785
text only 1,606
opponent 1,269
object or scene 990
other person 873
both 830
unknown 294
group crowd 232
Who is pictured
candidate 8,150
none 2,830
opponent 1,775
diverse crowd 962
veterans military 785
workers labor 676
family children 432
seniors 230
police 175
medical 121
faith figures 29
Candidate portrayal
warm relatable 4,757
not pictured 4,471
authoritative 3,235
working 163
unknown 145
heroic 108
Opponent portrayal
not pictured 9,075
unflattering negative 3,420
neutral 384
Symbols
none 7,146
us flag 2,418
farm rural 1,130
money cash 938
military 806
family 777
capitol statue 679
healthcare 434
factory industry 410
border 319
guns 2a 205
church cross 110
Setting
studio plain 3,362
outdoors nature 2,110
none 1,897
indoor home 1,800
official government 1,238
rural smalltown 703
unknown 688
workplace 539
urban 368
suburban 88
military 86
Palette
dark desaturated 3,218
patriotic rwb 3,136
neutral muted 3,034
warm bright 2,541
high contrast 882
unknown 68
Visual mood
angry urgent 3,046
calm trustworthy 3,007
energetic 2,937
hopeful uplifting 1,943
ominous threatening 1,181
neutral 593
somber 172
Text density
balanced 10,089
text heavy 2,631
image heavy 137
no text 22
Brand strength
minimal branding 9,985
strong logo branding 2,867
none 27
Borrowed credibility
none 9,147
news quote 1,638
poll stat 1,358
endorsement badge 968
as seen in 466
official seal 107
Authenticity
authentic candidate photo 7,789
unclear 2,319
screenshot real 2,055
doctored opponent 346
stock photo 329
likely ai generated 38
screenshot fabricated 3
AI / doctored flag
none 12,403
possibly doctored 419
possibly ai generated 57
📈 What's working — by production tier
By production tierpro vs DIY
valuecreativesavg days liveavg reach
professional 1742 11d 38,820
diy amateur 1876 12d 32,769
semi pro 9452 15d 28,156
unknown 4925 11d 11,236

Spend & reach are Ad-Library range midpoints. The Text and Visual analyses come from a Sonnet multimodal pass over each unique creative — classified once and shared across its duplicate ad copies, so those counts are per-creative.