Disinformation Report: August 1-15

How Anti-Trans Attacks Are Landing Across TFA’s Priority States — and How Campaigns Should Answer

American politics’ most infamous attack ad got an AI resurrection in Texas this period: a pro-Paxton super PAC remade LBJ’s 1964 “Daisy” spot with a generative-AI child on an operating table and “Dr. Talarico” behind the surgical mask — negligible reported paid spend, roughly 650K engagements of free amplification, and a campaign forced to answer a deep fake. Arizona, meanwhile, answered a month-old question the other way around: a candidate’s feed that reads silent while his campaign runs girls’-sports creative in paid rotation — persuasion that engagement tracking alone would never see. This briefing maps how the period’s attacks landed across the thirteen states TFA is tracking: who drove each story, who it reached, and where the openings are.

August 1 – 15, 2026

Data: Trans Rhetoric Social Media Dashboard

79K+ channels | 20 platforms | 1M+ posts

13 priority states | 711 state-tagged posts | 9.3M engagements | c4 / Organize Action

Topline Finding

Two states sit far clear of the field this period — Minnesota (~3.8M) and Texas (~3.1M) — and the reason each is there matters more than the rank. Minnesota tops the map because the national WNBA story physically staged itself in Minneapolis; most of that reach is national content with a local dateline (the narrative read is in the companion c3 briefing). The state’s genuinely electoral item is smaller and sharper: a 180K attack on Peggy Flanagan over gender-affirming care. Texas is there for electoral reasons all the way down: a pro-Paxton super PAC’s AI remake of the 1964 “Daisy” ad reached roughly 650K engagements on negligible reported spend because organic accounts — led by Libs of TikTok — did the distribution the PAC didn’t pay for, alongside the Texas Children’s settlement cluster and a 47–46 public poll. Below the top two, the pattern is candidate-shaped: Arizona’s standing Ciscomani question answered in the ad library rather than the feed, Iowa’s Democratic ticket repositioning under fire, a first candidate attack in Kansas, and a lawsuit by Nevada’s lieutenant governor in defense of his anti-trans task force. Maine — last period’s record-setter — collapsed to 108K: the attack machinery didn’t retire, it moved. Where it lands next is a scheduling question, not a hypothetical.

Source: Trans Rhetoric Dashboard | issue-location layer, off-topic virality removed | reach labeled for states above 100k

Minnesota and Texas run far clear of the field — one inflated by a national story’s local dateline, one electoral all the way down. Maine, last period’s record-setter, drops to the pack: attention follows the attack machinery, and the machinery moved.

Source: Trans Rhetoric Dashboard | issue-location layer, off-topic virality removed

Reach concentrates in Minnesota and Texas, with Wisconsin and Michigan a distant second tier and most states near zero. Concentration means the attacks are choosing their targets — not spraying.

Key Findings

• The AI attack ad has arrived, and it doesn’t need a media buy. A pro-Paxton super PAC remade LBJ’s “Daisy” ad with a generative-AI child on an operating table and “Dr. Talarico” behind the mask. Reported paid spend was negligible; organic amplification — Libs of TikTok’s clip alone at 567K, claiming the ad is “playing all over Texas” — delivered roughly 650K engagements free. Ad tracking that watches only spend will miss this format entirely; the creative is built to be laundered through engagement.

• Minnesota’s map-topping reach is mostly national content with a local dateline. The WNBA eligibility story staged itself at Target Center, and the state-tagged slice of that coverage (~1.2M) inflates Minnesota’s total — the narrative treatment belongs to the c3. What is electoral: the Senate Republicans account attacked Peggy Flanagan for supporting gender-affirming care (180K), and whatever the league eventually decides will land another national cycle in this state’s general-election window.

• Ciscomani answered the standing question — in paid, not organic. His account stayed quiet on girls’ sports for a second window, while his campaign’s “Out of Touch” spot runs a Tucson mother on sports and bathrooms — with a claim reporting has found false (Mendoza called the legislation “toxic,” not parents). Three of his six most recent spots run protect-girls terrain. Organic monitoring alone would have called this state silent.

• Texas’s parallel cluster is institutional medicine. The Texas Children’s Hospital whistleblower interview (816K) and “detransition clinic” coverage (92K), plus the Vance/RFK announcement of criminal referrals over gender-care billing (514K nationally), make hospitals and billing the emerging attack surface — stories with official actions attached, which age better than viral clips.

• Wisconsin’s biggest window closed with the candidate who anchored it. Rep. Tom Tiffany’s clip of a mother pressing Francesca Hong on the New Richmond bathroom situation (507K across two posts) attached the schools story to a named statewide candidate — but Hong is now out of the race, which retires the named target without retiring the format. New Richmond as a site, and the girls’-sports lane it anchors, carry into the general regardless of who tops the ticket.

• Maine’s collapse confirms the machinery is nomadic. From 255 posts and the largest state total on record to 21 posts and 108K in one window. The Webb attack apparatus didn’t retire — it moved to the WNBA story and the Texas ad. The trans-candidate protocol recommended last period is still the right posture; the next target gets chosen by the news cycle, not the calendar.

• The anti-trans audience got older and more male again. 72% male and Boomer-led this period (from 64% male, Gen-X-led) as the sports fight consolidated — while pro-trans reach nearly tripled (0.7M → 1.9M) on Whaley, Wade, and the sports counter-speech. The two sides are still in different rooms; both rooms grew.

The Battleground at a Glance

Two layers drive this briefing. The issue-location layer tags which state each post is about; the account-location layer tags where the messenger is based. A post can be about a state without originating there (a national account attacking Roy Cooper counts toward North Carolina), so the reach figures below measure attention on a state, not local activity alone. Off-topic virality that keyword-matched a state (foreign-language tabloid items, unrelated news) has been removed by hand.

State	Primary	Posts	Reach	vs last	Anti / Pro	Top messenger
Minnesota	8/11/26	156	3.8M	▲ 198%	202K / 517K	Fox News
Texas	TBD	201	3.1M	▲ 150%	517K / 1.1M	Catherine Herridge
Wisconsin	8/11/26	43	717K	▲ 212%	547K / 77K	Tom Tiffany
Michigan	8/4/26	136	658K	≈	80K / 170K	Gad Saad
Iowa	TBD	16	409K	▼ 42%	4K / 2K	David Weigel
North Carolina	3/3/26	38	170K	≈	167K / 251	America First Legal
Kansas	TBD	50	126K	▲ 748%	71K / 17K	New York Times Opini
Maine	6/9/26	19	108K	▼ 98%	7K / 21K	Ben Shapiro
Georgia	5/19/26	42	91K	▼ 30%	15K / 2K	Oshay Duke Jackson
Pennsylvania	5/19/26	13	81K	▲ 344%	215 / 3K	Fascinating
Arizona	8/4/26	21	44K	▲ 45%	5K / 38K	Mormon Stories Podca
New Hampshire	9/8/26	19	11K	▼ 93%	4K / 3K	Wisconsin Right Now
Nevada	6/9/26	10	4K	▼ 81%	583 / 905	Nevada Independent

“vs last” compares reach with the previous reporting period (≈ means within ±10%). A rising state deserves attention before its number is large.

Source: Trans Rhetoric Dashboard | stance classifier; "neutral" includes hostile-leaning news coverage

Tone is a directional read from the stance classifier; “neutral” absorbs a large amount of hostile-leaning coverage, so anti-trans shares are conservative — Minnesota is the clearest case this period: much of the Fox confrontation coverage codes neutral, so the state’s hostile share understates the reach’s actual valence. Texas has the same problem in reverse — the Daisy-ad debunk coverage (Mother Jones, LGBTQ Nation) also codes neutral. Read the stance chart alongside the notes, not instead of them.

How the Attacks Travel — Originators vs. Amplifiers

Two different jobs create anti-trans reach. A small set of accounts originate stories — surfacing a local incident that then spreads; a larger set amplify, putting the most eyeballs on stories someone else started. The two call for different campaign responses: pre-empt originators, rapid-respond to amplifiers.

Top originators (seed stories that travel): Fox News (32.7M, 33 seed stories); Matt Walsh (6.1M, 2 seed stories); OutKick (4.5M, 6 seed stories); WISH-TV News (3.6M, 2 seed stories); Royce White (3.2M, 10 seed stories).

Top amplifiers (most reach on others’ stories): Fox News (all accounts) (26.1M, 49 reposts); NEWSMAX (all accounts) (823K, 21 reposts); Visegrád 24 (657K, 4 reposts); Rep. Nancy Mace (526K, 5 reposts).

Who These Attacks Reach

The dashboard has no engager-level data, so audience is modeled — inferred from the demographics of the accounts that produce and amplify content, weighted by engagement, under a homophily assumption. It blends US Census surname data, Social Security first-name records, each platform’s known user base, and where accounts are based. For the roughly 82% of anti-trans reach carried by accounts with a real name or seed match this period, this is a genuine read; the rest falls back to a national baseline. Treat it as directional, not measured — but the split between the two sides is stark enough to act on.

Basis notes: nine Texas AI-ad posts (incl. the Libs of TikTok amplification) were hand-tagged to Texas after the state classifier missed them; the diffusion pass ran August 17 over a window extending slightly past the period close (role rankings unaffected).

Audience of…	Gender	Top age cohorts	Race (modeled)	Evidence	Conf.
Anti-trans attacks	72% male / 28% female	Boomer 29% · Gen X 26% · Millennial 24%	White 78% · Black 9% · Hispanic 6%	79%	0.71
Pro-trans content	62% female / 38% male	Millennial 30% · Gen X 27% · Boomer 24%	White 87% · Hispanic 6% · Black 5%	66%	0.59
All state content	58% male / 42% female	Millennial 32% · Gen X 26% · Boomer 22%	White 79% · Hispanic 9% · Black 7%	60%	0.58

The masculinization continued: anti-trans state content reached a 72%-male, Boomer-led audience this period (from 64% male, Gen-X-led last period) as the sports fight consolidated through Fox and sports media. The countervailing move is on the other row: pro-trans reach nearly tripled (0.7M → 1.9M) on the Whaley and Wade candidate stories and the sports counter-speech, reaching its usual majority-female millennial-to-Gen-X audience. The rooms are still different rooms — but for the first time in three periods, both grew. Measurement note: the evidence share stays high on the anti-trans row (79%) with confidence at 0.71 — again because the reach ran through large named outlets — and rose on the pro side (66%). “Evidence” is the share of reach from accounts with a real name/seed match rather than a national-average fallback; “Conf.” is the mean modeled confidence. The racial estimate is the weakest layer and should be read most loosely.

Case Studies

Case Study 1: Texas — The AI Resurrection of the Daisy Ad

What happened. In early August, a super PAC supporting Ken Paxton’s Senate campaign released a 30-second, fully AI-generated remake of LBJ’s 1964 “Daisy” ad. The original’s child in a field counts petals; this one’s child moves to a hospital bed, an oxygen mask, and a surgeon who pulls down his mask to reveal an AI-generated James Talarico — nuclear war swapped for a fabricated gender surgery on a minor. Talarico’s campaign told Mother Jones that “James does not support gender reassignment surgery for minors” and that “this AI deep fake ad is flat out lying”; such surgeries are rarely if ever performed on minors. Reported paid spend behind the spot was negligible — but organic distribution was not: Libs of TikTok posted the creative to 567K engagements with the caption “this ad is playing all over Texas,” Ryan Girdusky praised the craft to 67K, and the debunk coverage (Mother Jones, LGBTQ Nation) added roughly 25K. Total: ~650K engagements the PAC didn’t buy.

Why it matters. This is a new economics for attack ads. The creative is not a media buy with organic spillover — it is an engagement product wearing an ad’s clothes, released at token spend and distributed by the amplification network for free. AdImpact-style spend tracking reads this as a non-event; engagement tracking reads it as one of the period’s larger Texas stories. It is also the first fully generative attack creative in a tracked priority race — the June Talarico-in-a-dress AI spot was the prototype; the Daisy remake is the format maturing: borrowed prestige, fabricated imagery, deniable scale.

Takeaway. Track creative, not just spend — the watchlist now flags this ad’s distribution chain, and the ad-library check belongs in every cycle alongside the engagement pull. The response side has a working receipt: the campaign’s own denial plus the clinical fact (minors do not receive these surgeries) is the accurate chain, and it should be pre-packaged for the next AI resurrection — this format is cheap enough that there will be one.

Case Study 2: Iowa — Repositioning Under Fire

What happened. The period’s most-shared Iowa item was not an attack ad — it was David Weigel’s widely shared read (281K) of a Democratic candidate in a Trump+4 seat declaring she “oppose[s] letting men in women’s locker rooms.” Republicans attacked her anyway, pointing to her vote against a 2023 omnibus GOP gender bill, and Weigel noted the statewide ticket — Sand and Turek both — has moved the same direction. As he framed the dynamic trans-rights groups had predicted: change your position and you’ll get hit anyway.

Why it matters. This is the clearest documented case in the tracking of Democratic repositioning under pressure in a priority state — and of its limits. The reposition bought the candidate no ceasefire: the attack simply shifted from her position to her record. For every campaign weighing the same move, Iowa is the natural experiment, and the early result is that the attacks are about the fight, not the stance — a candidate who concedes the frame still carries the record, and the record is what gets run against her.

Takeaway. The strategic question a campaign should ask before repositioning is not “does the new position poll better?” but “does the attack stop?” — and Iowa’s evidence says no. That cuts toward answering on disproportion and record-defense grounds rather than chasing the position: the vote will be litigated either way, and the candidate who has conceded the frame defends it from weaker ground. Worth watching whether the Iowa ticket’s numbers move at all in response — the reposition’s actual electoral yield is the open empirical question.

Case Study 3: Arizona — The Quiet Feed and the Loud Ad Library

What happened. For a second consecutive window, Rep. Juan Ciscomani’s accounts posted nothing on girls’ sports — 21 state-tagged Arizona posts and 44K reach would read as a dormant state. The ad-library check (the standing July 8 request) found the opposite: his campaign’s “Out of Touch” spot features a Tucson mother saying of JoAnna Mendoza that “she even wants guys in girls’ bathrooms” and claiming Mendoza called parents like her “toxic.” Reporting found the toxic claim false — Mendoza called the legislation “harmful… toxic,” and never applied the word to parents. The same reporting counts three of Ciscomani’s six most recent spots on protect-girls terrain, in a race a recent poll has Mendoza leading by two. Whether the specific July creative from the original request is among them is not established by the public reporting — next cycle’s ad-library pull should enumerate the full set. In parallel, the Prop 318 school-sports initiative produced its first creative: the Yes campaign’s “Pretending” video.

Why it matters. This is the split the July 8 ask anticipated: organic silence with paid escalation. A campaign that has learned what the engagement trackers see can run the trans attack entirely in paid — targeted, unshareable, invisible to feed monitoring — while its candidate’s accounts stay clean. Arizona now has the inverse profile of Minnesota: near-zero engagement, active paid persuasion. Whatever the format’s spread, the monitoring lesson generalizes to every priority race.

Takeaway. The ad-library check graduates from a one-off request to a per-cycle step, and quiet states get it first, not last. The response material writes itself from the record: the full Mendoza quote — legislation, not parents — is the receipt, and the misquote is the vulnerability. On the ballot fight, Prop 318 messaging is untested terrain — the modeled message guidance in Strategic Recommendations is the starting hypothesis, not a result.

Attack Formats to Watch

• The AI-resurrection ad. Texas’s Daisy remake: prestige creative fabricated by generative AI, released at token spend, distributed free by the amplification network. Spend-based tracking reads it as a non-event. The counter is the pre-packaged receipt (the campaign denial plus the clinical facts) and creative-level tracking — this format is cheap enough to recur before November.

• The paid-quiet split. Arizona’s lesson: organic silence while the attack runs entirely in paid rotation, targeted and invisible to feed monitoring. Quiet states get the ad-library check first. The tell is a candidate whose accounts have gone unusually clean on the issue his ads are running on.

• The false-paraphrase ad. Arizona’s “Out of Touch” spot puts a fabricated quote-shape in a real voter’s mouth: Mendoza called the legislation toxic; the ad says she called parents toxic. The paraphrase is the vulnerability — it is checkably false, the full quote is the receipt, and a documented false claim in paid creative is actionable in ways organic distortion is not.

• The institutional-medicine story. The Texas Children’s Hospital whistleblower interview, the settlement’s “detransition clinic” terms, and the criminal-referrals announcement over gender-care billing: attacks with official actions attached, which produce document-driven follow-ups for months and age better than viral clips. Hospitals and billing codes are the new school boards.

• The hit-them-anyway trap. Iowa’s lesson, run as a format: when a Democrat repositions on trans issues, the attack does not stop — it moves to the record. The reposition is treated as an admission and the old vote as the proof. Campaigns weighing a position change should price in that the fight continues either way, from weaker ground.

Looking Ahead — The Primary Calendar

October 13 is the next hard date on the calendar. The Medicaid rule takes effect mid-general-election, and the decision moves to the states — every priority-state campaign will be asked whether their state should pick up coverage. The “may, not must” receipt should be in campaigns’ hands before the date — and the modeled message guidance for this terrain sits in Strategic Recommendations, labeled as the prediction it is.

Watch for the Daisy remake to acquire real money — or successors. The format’s economics (token spend, free distribution) make copycats likely in other Senate races; the creative itself may yet enter genuine paid rotation in Texas. The watchlist now flags the distribution chain, and the ad-library check runs every cycle — quiet states first, per Arizona’s lesson.

The institutional-medicine docket grows through the fall. The Texas Children’s settlement (its “detransition clinic” must open by late October), the gender-care billing referrals, and AFL v. Cabarrus County (now with a companion records suit) will each produce document-driven follow-ups for months — official actions are attack surfaces or openings depending on who frames the documents first.

One primary remains: New Hampshire, September 8. The last competitive-primary window in the set, with SB 430 implementation stories arriving with the school year and the Wade story establishing candidate-level stakes.

Strategic Recommendations

COUNTER	PIVOT AWAY	LEAN INTO
• The AI-ad receipt (Texas): the pre-packaged accurate chain for the Daisy remake and its successors — the campaign’s denial, the clinical fact that these surgeries are rarely if ever performed on minors, and the ad’s fabricated provenance — ready before the next resurrection. • The misquote correction (Arizona): Mendoza called the legislation toxic, not parents — the full quote is the receipt, and the false paraphrase in paid creative is the documented vulnerability.	• Repositioning in search of a ceasefire — Iowa’s evidence is that the attack continues from weaker ground; the record gets litigated either way. • On care terrain (modeled prediction — see note below): slippery-slope/“erasure” framing, prosecute-the-doctors outrage, and leaning on medical authority all test negative; the family reads as the trusted authority, not the institution.	• The per-cycle ad-library check, quiet states first — Arizona proved the feed can read silent while the persuasion runs in paid; creative-level tracking is now part of the watch. • On care terrain (modeled prediction — see note below): the practical, government-in-the-exam-room story with the family-cost mechanics — the October 13 date, who stops paying, the state’s choice — with “the government,” not “politicians in Washington,” as the named actor.

A note on the message guidance above. The care-terrain items come from the August 11 MaxDiff panel on the Medicaid rule. Since the Tavern feed ended in June there is no live survey data calibrating that panel, so its results are predictions from a synthetic model, not measurements — directionally useful as a starting hypothesis, and no substitute for live testing before any ad spend or scripted use. The strongest modeled results, for the record: the exam-room overreach message (76% overall / 89% Independents), the family-cost and effective-date messages (85%+ with Independents), practical register over empathetic, “the government” over “politicians in Washington” — and erasure framing, doctor-criminalization outrage, and medical-authority appeals all modeled as hurting. Treat every number in this paragraph as a hypothesis with a test plan attached.

The bottom line. The period’s two defining electoral stories are both about invisibility to the old monitoring: an AI attack ad that needed no media buy, and a paid campaign that needed no posts. Iowa adds the third lesson — repositioning buys no ceasefire. Build the two receipts (the AI-ad chain and the Mendoza quote), stand up the per-cycle ad-library check before the fall creative arrives, and put the modeled care-terrain guidance into live testing while the October 13 clock is still running.

Limitations

Geography is measured from two dashboard layers: issue-location (which state a post is about; ~10% of posts carry a state tag) and account-location (where a channel is based; ~2% of channels). A post can be about a state without originating there, so state reach measures attention, not local activity. Off-topic virality that keyword-matched a state (foreign-language tabloid items, unrelated news) was removed by hand. Stance (anti/pro/neutral) is a directional classifier read; “neutral” absorbs much hostile-leaning news, so anti-trans shares are conservative.

The audience read is modeled, not observed: the dashboard has no engager-level data, so audience is inferred from the demographics of the accounts producing and amplifying content, engagement-weighted, under a homophily assumption (producer ≠ audience is the foundational caveat). It blends Census surname data, SSA first-name records, platform user base, and account location; the “evidence” column reports the share of reach with a real name/seed match versus a national fallback. The racial estimate is the weakest layer (only ~2% of channels have a state) and is hedged hardest. The originator/amplifier read is modeled from near-duplicate text clustering plus native amplifier edges captured for ~36% of posts. The MaxDif guidance in Strategic Recommendations is modeled × modeled — a synthetic panel with no live calibration since the Tavern feed ended — reported there as a prediction only. Engagement is not persuasion.

Methodology

Data source. Trans Rhetoric Social Media Dashboard, built for TFA on the Junkipedia social listening API — 79,000+ channels across 20 platforms, 1,000,000+ posts spanning December 2023 to present, collected on 90+ keywords (topic terms, framing terms, and named trans public figures). Period: August 1 – 15, 2026. Priority states are drawn from Adrienne’s “TFA priority states-races 5.18.26” sheet, plus Iowa (Jun 9 request) and Texas/Talarico. Per-state signal uses two geolocation layers — issue-location (which state a post concerns) and account-location (where a channel is based) — filtered to trans-relevant posts carrying engagement, with a manual audit of the high-engagement tail to remove off-topic keyword matches.

Diffusion: Originators vs. Amplifiers

The originator/amplifier analysis separates the accounts that start a story from those that give it the most reach. It runs in two passes. First, near-duplicate clustering: every post’s text is normalized (URLs, @-handles, casing, and punctuation stripped) and grouped with near-identical posts, so the same claim or clip is tracked as it reappears across accounts — a window’s posts collapse into a few thousand recurring-story clusters covering roughly 40% of volume. Second, native amplifier edges: the platform-level quote, reply, and repost relationships (present on about 36% of posts) link each amplifying post to the item it echoes.

Within a cluster, the originator is the earliest poster (or, for a retweet/quote chain, the account being quoted); downstream reach is the total engagement the story accrues beyond the originator’s own post; amplifiers are the other accounts in the cluster, ranked by the reach they add to stories they did not start. Originators are ranked both by the number of distinct stories they seed and by the downstream reach those seeds generate. Caveats: native edges exist on ~36% of posts; clustering misses heavy paraphrase; and “originator” is the earliest post in the dataset, which can trail an off-platform origin.

Audience Model

Because the dashboard captures no engager-level data, audience is modeled from the accounts that produce and amplify each slice of content, engagement-weighted, under a homophily assumption (producer ≠ audience is the foundational caveat), blending Census, SSA, platform, and location data. Legal characterizations of the Medicaid rule follow the published rule: federal funding ends, state-only coverage is permitted, a six-month taper is allowed — “may, not must.”