For years, Texas Republicans and their allies have been quietly rigging the systems that decide what you find when you go looking — search results, AI answers, Wikipedia, the “official” sources underneath all of it. Nobody sees the manipulation. They just see the answer.
It will take citizen power to unf*ck it. That means you.
Some sources say widespread fraud is a documented problem, citing a government-hosted database of cases…
assembled from planted sources
████.gov › hearing-materials
Election Fraud Case Database (434 pp.)
think-tank doc, .gov authority
thelonestarledger.example › local
Local officials “alarmed” by new voting numbers
looks local. isn’t.
aggregator.example › trending
Same claim, 40 “independent” headlines
one source, copied 40×
01
This is bigger than misinformation.
Misinformation is a symptom. What’s actually happening is manipulation of the machinery that decides what people find.
The old fight
A false post goes viral.
You see it. You correct it. You move on. The battleground is the feed, and the timeline is days.
The fight now
Someone asks a question.
They ask Google, or ChatGPT, or an AI summary. The answer is assembled from whatever sources exist — or whatever sources have been planted and amplified. The manipulation happened months ago, upstream, where nobody was watching.
They don’t need to win the argument. They just need to own the answer.
02
The recipe.
Six ingredients show up again and again in the Republican information architecture in Texas. None of them are improvised. They’ve been building this for years.
Manufactured Consensus
Serves: whoever’s searchingPrep: yearsCook: the week before you vote
Semantic interference
Capture the definitions. Attach their language to your name, your district, your issue — so when people look you up, they find their framing first.
.gov authority laundering
Publish partisan material on government domains, which machines treat as authoritative by default. A conservative think tank’s 434-page election-fraud compilation, posted to Congress.gov as hearing material, ranks for roughly 192 election-related searches.
Aggregator amplification
Feed news aggregators and syndication networks so one claim appears as dozens of “independent” sources. Repetition reads as corroboration.
Pink slime
Sites built to look like local news, filling the space real local papers left behind. In news deserts, they are the local news.
Targeting the AI layer
Deliberately build the sources models retrieve and cite. In an audit of opposition-aligned groups, four of five had 60–99% of their reference citations in non-English Wikipedias — invisible to English-only watchers, fully visible to AI.
Kernel-of-truth seeding
Attach the whole operation to something real and broadly agreed on, so the package inherits credibility. Expect them wherever there’s real consensus. That’s the point.
Secret ingredient
Data voids.
When search engines and AI have nothing good to return, whoever fills the void controls the frame. Silence isn’t neutral. Silence is how data voids form.
03
How a belief gets cooked.
People do exactly what media literacy taught them: they go look it up. The trap is waiting at every step. Tap each stage.
What you experience
A moment of curiosity, doubt, or confusion. You become a seeker.
What they engineer
Trigger seeding. Ads, posts, speeches, and coordinated amplification time the doubt — so it lands right where they’ve already prepared the answer.
What you experience
You search. You ask an AI. You check across sources. You do everything right.
What they engineer
Query shaping. Coined terms and prepared data voids decide which words you type — and those exact words lead straight to their pages.
What you experience
The results agree with each other. The attitude sets — and resists correction.
What they engineer
Result capture. Official-looking sources supply manufactured corroboration at the top of the page. Each cycle narrows the next query.
Research base: Aslett et al., Nature (2024) — searching to evaluate false news can increase belief in it · Golebiewski & boyd, Data & Society (2019) · Tripodi, Data & Society (2018) · Epstein & Robertson, PNAS (2015).
04
Texas is on the menu.
Texans are looking things up at a pace we’ve never seen. What they find is being decided right now.
7,348
views in one day on Wikipedia’s 2026 Texas elections article — the day after the March primary. The biggest spike in eight years of Texas election data.
2.4×
more Wikipedia attention in 2026’s primary week than 2022’s — same article type, same point in the cycle.
6.3%
of what Texans were paying attention to online in one June snapshot was politics. The rest of the feed is where opinions actually get shaped.
528,086
backlinks from 13,060 domains feed the digital authority of the Texas Governor’s office — the scaffolding search and AI lean on.
Sources: Wikipedia Pageviews API (all agents), en.wikipedia.org; Texas online-attention snapshot, June 24, 2026; backlink index pull, June 2026.
The empty shelf.
Some of the biggest issues in Texas don’t have a Wikipedia article at all. When nothing neutral exists, AI and search fill the gap with whatever’s loudest.
Property taxes in Texas
Cost of living in Texas
School vouchers in Texas
Medicaid in Texas
Housing in Texas
Texas election law
Political misinformation in Texas
Border Patrol in Texas
Foreign ownership of land in Texas
Each of these titles returned “Wikipedia does not have an article with this exact name” when checked on September 25, 2026.
Why it takes us
Billionaires own the algorithms. Official offices can manufacture headlines on command. The opposition has money, power, and a head start.
What we have
The truth, and numbers. There are more of us than them — and regular people are the most trusted messengers there are.
05
Build the digital firewall.
Not one campaign — a stack, built in layers. Not volume: interception. Be present, together, at the moment people are already looking for answers.
THE MOMENT
The moment. People already looking for answers. The point of interception.
Accurate base. Facts, records, sources. The standing advantage we already hold.
Translation layer. Creators, podcasts, videos, neighbors — making accurate information fit the moment.
Drive demand. Give people reasons to go looking — and make sure the truth is there when they do.
Coordinate. Organized by role, networked, sharing what we learn. The structure that holds it all together.
Rules of the firewall
01
Be the sandbag.
Inoculate your people first. Prebunk what’s coming before it lands.
02
Correct the record. Starve their account.
Don’t like, comment on, quote, or link their content — even to dunk. Use screenshots or archive links.
03
Stop being quiet.
Silence is how data voids form. If we don’t fill the space, they will.
04
Your own words, always.
Copy-paste toolkits collapse the message and look like coordinated inauthentic behavior. Say it like you say it.
05
Through process, not force.
Every platform has a way to fix things. Use it. Edit wars and mass-reporting lose — and hand them a story.
06
Follow the people, not the elites.
AI answers cite Reddit, YouTube, and news more than encyclopedias. Show up where people are actually talking.
06
Resisting algorithmic manipulation: a field guide.
Name it correctly and the response follows.
Tap a card for the response.
Red-flag check
One flag isn’t proof. If a post trips several at once, treat it as designed.
Check what you see.
Truth-sandwich builder
Lead with the fact. Name the myth once, briefly. Close on the fact. (Concept: George Lakoff.)
Before you believe or share
Source
Who created or shared this? Are they named and credible?
Intent
Why is it framed this way?
Context
What’s missing, exaggerated, or cherry-picked?
Benefit
Who gains if I believe or share this?
Verify
Can independent, trusted sources confirm it?
Ethics
Is it safe, legal, and responsible to share?
The golden rule: don’t amplify false or harmful content — even to correct it.
07
Unf*ck the algorithms.
Pick one mission that fits the time you actually have. Small, real, repeated actions are what move the machine.
Fix the record on Wikipedia.
Wikipedia is one of the most heavily weighted sources behind search results, knowledge panels, and AI answers. It’s also where scandals quietly get softened, sourced facts get trimmed, and entire issues go missing. Anyone can help keep it complete and accurate — for James Talarico, for Ken Paxton and every other candidate, and for the Texas issues people are searching.
Read before you edit
Accuracy, not advocacy. Add what reliable, independent sources say — court records, major papers, the Texas Tribune. Never your opinion, never campaign material.
Talk page first. For anything contested, propose the change on the article’s Talk page with sources. Wait for consensus.
Never edit war. If you’re reverted, discuss. Reverting back turns a content dispute into a case against you.
Disclose conflicts. If you work for or are paid by a campaign, party, or candidate, say so on your user page and use “edit request” on the Talk page instead of editing directly.
Go solo. Use your own account, your own words, your own judgment. Coordinated or copy-pasted edits get pages locked and editors banned.
Watch & protect
James Talarico
High-traffic pages attract vandalism and attack edits as elections near. Watchlist it. Revert only obvious vandalism; take everything else to the Talk page with sources.
The impeachment, the securities-fraud case and how it ended, the whistleblower lawsuit — make sure the documented record stays reliably sourced and isn’t quietly trimmed. Check the history tab for removals of sourced material.
Nine big Texas issues have no article at all. Gather independent sources and draft one through Articles for Creation — reviewers help new writers get it right.
One report is a complaint. A log of hundreds is evidence. Walk through the steps below: capture it without amplifying it, report it through the platform’s own process, then drop it into the shared, anonymous log so the pattern gets documented.
Honest expectations
Report real violations only. Platforms act on rules, not on disagreement. Pick the rule it breaks.
One honest report each. Mass-reporting and brigading break platform rules and discredit real reports.
Never engage the post. No replies, likes, or quote-posts. That feeds the algorithm.
Keep yourself out of it. Don’t put your name or personal details in the log.
Fill the voids. Build the durable record.
The fastest way to beat a cooked answer is to make sure a better one exists — somewhere machines trust, in your own words. Pick one bucket that fits the capacity you actually have this month.
Durable record
Permanent, indexed, heavily weighted
Public comment on regulatory dockets. Official, searchable, and it lives on .gov domains too.
Legislative Reference Library. The LRL takes submissions — and already takes partisan outlets. Send well-sourced reporting and newsletters.
Letters and op-eds in your local and regional papers.
Wikipedia and Wikidata, via the Talk page and reliable sources.
Your own blog, newsletter, or Substack. Archives get indexed.
Answer layer
What AI actually retrieves and cites
YouTube. Where political search goes: debate clips, speeches, explainers. Titles and descriptions that answer the real question.
Reddit. Genuine participation in your city and issue subs — not drive-by links.
Long-form writing. Substack, Medium, community blogs.
Forums where your issue already lives. Parent groups, hobby boards, neighborhood apps.
In the moment
Interrupt it while it forms
Community Notes on X and Meta’s apps. Notes need raters from different perspectives — write for them.
The comment layer on high-visibility news pieces, with sources.
Trend-jack and pivot. Join the conversation people are already having, then point to the record.
Talk to your people. A text to five friends beats a post to five thousand strangers.
Documentation
Build the pattern
The incident log. Mission B. Every entry makes the next report stronger.
Screenshot the search. When a Texas search or AI answer looks cooked, capture the query, date, and results.
Platform reporting, with honest expectations.
None of this matters if we don’t vote.
The deadline to register in Texas is Monday, October 5. Early voting runs October 19–30. Election Day is Tuesday, November 3.
False claims about when, where, or how to vote are a special case: report them to the platform and to the Texas Secretary of State Elections Division at 1-800-252-VOTE (8683).
Then text three people you know and ask if they’re registered. That’s the whole mission.
Send this to the one person who needs it.
Not a thousand strangers. One person you know. That’s how this spreads.