Right, so Israel's propaganda operation told everyone to check their facts about Gaza, and, unsurprisingly, checking those facts leads you right back to Israel's own state-funded propaganda. It’s like handing the burglar the keys to the evidence room, then acting surprised when the trail ends at his own front door. Bibi the Burglar apparently.
Because when investigators actually did what Bibi suggested, check those receipts, demand truth, they found themselves in a digital network of websites made not for people, but for search engines, web crawlers, and AI bots. Fill them with search-optimised nonsense, get them indexed, and when someone asks a chatbot about Gaza, one of those pages might show up as a 'source.' No secret back door into ChatGPT is needed. Just stock the shelves and hope the robot librarian picks your book of propaganda first. The trail led through Brad Parscale’s Clock Tower X, the US political-media outfit hired to do the work, then through Havas Media Germany, the intermediary named in the foreign-agent filing, and back to the State of Israel on whose behalf the campaign was being run. That matters because these were not random websites with a suspiciously familiar line. They were part of a documented state-backed operation.
They said to follow the source. So people did.
Bad advice, Bibi.
But don't just take my word for it. Listen to how investigator Nick Cleveland-Stout described what these sites were built to do.
Traditional propaganda tries to persuade you directly. It buys ads, pushes videos into your feed, hires some too easily bought and paid for influencer wearing too much slap or with stupid hair to explain why the thing you just saw didn't really happen, and then calls the resulting outrage an outbreak of antisemitism.
Well this operation seemingly tried something more covert, like propaganda with a silencer. Why shout your message when you can whisper it to a robot and let it spread for you, like digital poison?
Build the websites. Fill them with organised content. Optimise them for search. Get them indexed. Then wait for an AI system to go searching for sources.
Suppose someone asks a web-connected chatbot, "Is Israel blocking aid into Gaza?" The chatbot searches online and might pull in one of those campaign sites alongside regular news reports. The user gets a tidy summary, but unless they check every source and scroll down to the disclosure, they might never realise one was paid for by the very state they were asking about.
These websites didn't need to win you over. They just needed to charm the robot you ask for homework help. The real audience isn't you; it's your digital assistant, and apparently that's easier to fool than a government minister with trip to Tel Aviv, or you are these days with more traditional hasbara.
But before anyone rushes off to buy a tinfoil hat here, no, this is not proof that Israel hacked ChatGPT. Netanyahu is not sitting in a Tel Aviv basement, finger over a big red button labelled 'make robot more Zionist.'
It does not prove that Israel secretly retrained OpenAI's models. It does not prove it controls Claude, Gemini, Copilot, or every chatbot answer someone receives.
And saying that would be pretty stupid actually, because it gives Israel's defenders the world's easiest escape: knock down the wild exaggeration, ignore the real evidence, and call the whole thing fake news.
What the public record shows is narrower, more specific, but still deeply troubling.
A public US foreign-agent filing from Clock Tower X described work through Havas Media Germany for the State of Israel. That work included mass content production, search-engine optimisation, and, in the filing's exact words:
"Deployment of websites and content to deliver GPT framing results on GPT conversations."
Not model ownership. Not a secret back door.
A plan to arrange the library shelves before the robot librarian is aware their carefully curated system has been defiled. It’s a bit like hiding all the vegetables behind the chocolate bars and then acting surprised when the robot stats recommending diabetes as a lifestyle choice.
The initial field work plan set out a combined budget of $1.5 million a month for four months-six million dollars in that first pool.
That's not the same as the later $46.5 million figure reported by Drop Site, but I’ll come onto that in a minute.
The first filing also called for at least 100 root creative assets each month, 5,000 variations, 80% of the material aimed at Gen Z, and a minimum of 50 million impressions a month.
So the GPT language wasn't just a stray typo squeezed between a pitch for three Facebook posts and a free novelty mug.
This was messaging on an industrial scale, aimed at exactly the audience Netanyahu says Israel is losing.
And he has already told us how he sees that audience.
Because seven fronts just weren't enough for Bibi. He had to open a new front against teenagers with TikTok accounts, against young people with phones, too. Who would have guessed that would be Israel’s next war front?
People watch Gaza being bombed, see families displaced, hear allegations of starvation and collective punishment, and still refuse to give Bibi the standing ovation he thinks he deserves. The nerve of the public. How dare they not applaud his war crimes(!)
So, in Netanyahu's mind, they haven't reached a political judgment. They have just become enemy territory.
And once you start treating public opinion as a battlefield, criticism becomes enemy fire doesn’t it? Journalism becomes a hostile weapon. A university protest becomes an attack. An inconvenient photograph becomes disinformation. A young person refusing to accept the Israeli government's version becomes a target to be won back, suppressed, or outnumbered by algorithms.
Because the blame can never be with the policy, the blame must be with the marketing.
Maybe the logo was too small. Maybe the slogan needed more exclamation marks. Maybe Gaza would look less like a war crime if Brad Parscale just churned out 5,000 slightly altered websites telling you not to trust your own eyeballs. If only reality could be fixed with a catchy jingle and Netanyahu dressed as a 1970’s game show host.
But Netanyahu was even more direct in his opining on this elsewhere.
"The most important ones are on social media."
Most of us call them apps. Bibi calls them weapons. At least he's finally being honest about something. Next he'll be calling TikTok a WMD and demanding sanctions on ASMR creators.
Clock Tower X is the next step beyond flooding someone's social feed.
Don't just put the message in front of the user. Place sources behind the system the user consults.
It's like stocking a supermarket shelf before the customer arrives. You don't need to own the supermarket. You don't need to bribe the cashier. You make sure your product is everywhere, facing forward, right at eye level, and wrapped in packaging meant to look more trustworthy than it really is.
Except here, the customer is an automated retrieval system, and the product is state-funded framing about Gaza.
Drop Site reported that ten sites linked to the campaign appeared hundreds of times in Common Crawl between January and June.
Common Crawl is a non-profit web archive that automatically crawls and stores copies of public webpages at huge scale, creating huge datasets AI developers can use to build models. So appearing there means a page entered a widely used pool of web material. It does not prove that any specific chatbot was trained on it or that it influenced a particular answer.
And that’s an important difference. An archive is a possible route, not proof of arrival.
Similarly, one chatbot citing one campaign-linked website does not prove the campaign caused a certain answer to be created. Results can change depending on the model, the wording of the question, the date, the user's location, and whether live web search is enabled.
But none of that erases the work plan.
Israel paid for an operation meant to influence GPT-style framing. The network existed. The sites were found. And at least one apparently ordinary source turned out to be much less ordinary than it first seemed.
Allyvia. Sounds like a wellness app that sends you reminders to hydrate. Instead, investigators identified it as one of the sites tied to Parscale's campaign. So much for digital self-care; this one's more about political self-delusion.
And this is where the credibility operation starts to pollute its own product.
A chatbot user sees a website name under an answer and assumes it's a regular publication. Don't. Open the source. Check its About page, contact details, the footer for T’s and C’s. That kind of thing. Search the site's name with words like 'owner,' 'funding,' and 'foreign agent,' then compare its claims with truly independent reporting. A polished logo only proves someone knows how to use Photoshop, not that they're telling the truth.
In this case, scrolling to the footer showed that Clock Tower X distributed the material for the State of Israel-a rather important detail for the chatbot to leave out and of course I’ve covered such shenanigans by Clock Tower X on this channel before, election laws and the like requiring who you’ve acted on behalf of to be printed clearly, even if you do try and make it as miniscule as possible.
Most people won't do that though.
That's why they asked the chatbot. They wanted the short answer, not a forty-minute audit of every source and its funding.
The disclosure might technically exist on the original webpage while disappearing from the experience of getting the answer.
That does not automatically make every factual statement on the site false, but it does mean the speaker's identity is hidden at the exact moment credibility is transferred.
A page paid for by a foreign state is not independent evidence just because a chatbot put it next to other sources and spoke in the calm voice of an automated teaching assistant.
And once the network is exposed, the appearance of independence starts to fall apart.
Ten different sites no longer look like ten different voices. They start to look like one campaign wearing ten hats and hoping nobody notices they're all sitting on Brad Parscale's head. It's the digital version of a bad disguise: a fake moustache, fake glasses, but the same old propaganda.
But the dramatic headline is not the full story either. Testing threw up some surprises.
ChatGPT and Claude were reportedly more resistant to repeating the campaign's framing during those tests.
Now far from weakening the story, this is rather useful, because it keeps things from turning into propaganda about propaganda.
It makes the honest position a lot more simple too.
The filing proves intent and infrastructure. The investigation reportedly found campaign-linked sources appearing in certain tests. Common Crawl shows a possible route into the wider information ecosystem.
And yet none of that proves universal success, because some GPTs were more resistant to their nonsense than others.
Israel tried to influence the source environment, but clearly they cannot control the machines quite yet.
And it means the real human consequence here is not someone’s algorithm becoming compromised.
It is you.
It is anyone asking a complex question and trusting the answer to reveal its sources.
Most users will not check who funded each website. They will not inspect corporate registrations, foreign-agent filings, and footer disclosures before asking what happened in Gaza.
They will see a concise answer delivered with machine confidence.
And machine confidence is very useful to propagandists. It can make a political argument sound like a neutral summary. It can strip away the context of who happened to pay for a given source, or why the page exists, and which audience the campaign wanted to influence. The trick is to turn propaganda into homework help.
The danger is not that a chatbot reaches through the screen and forces you to support Israel.
The danger is that paid state messaging can adopt the voice of a supposedly neutral machine while its political identity disappears somewhere between the source page and the answer.
And it will not stop with Israel.
This broader tactic is not a new one. Governments, corporations, and PR firms already use search optimisation, sponsored articles, front groups, and content farms to crowd out critical results and create the illusion of consensus. The AI twist is making that material easy for chatbots to find and cite.
If flooding the source environment works there, every military contractor, fossil-fuel company, billionaire lobby, and would-be autocrat will want to try it too. The line for digital manipulation is about to get longer than the queue for free money at a Davos buffet.
Why persuade the public one by one when you can try to influence the systems millions of people consult?
That is why calling this ordinary public diplomacy is too soft.
The paperwork describes the wider campaign's stated purpose as combating antisemitism in the United States. That should be reported fairly. Combating antisemitism – the real thing - is a legitimate and necessary goal.
But that respectable purpose does not erase the rest of the work plan.
The same plan described mass narrative production, search visibility, targeting younger people particularly, and websites meant to affect GPT framing.
Criticism of Israel's actions is not automatically antisemitism, the conflation is itself antisemitism. Disbelief in Israeli government messaging is not hatred of Jewish people, no matter how many Israeli ministers go on TV, stamp their tiny feet, turn on the water works and plead victimhood like their life depends on it. Yet blurring those lines gives the campaign something handy to hide behind too.
Challenge the government's version, and suddenly the argument is no longer about what Israel did. It becomes an argument about what kind of person you supposedly are for noticing.
Policy becomes messaging.
Opposition becomes prejudice.
Falling credibility becomes a technical problem needing more money, more websites, and thousands more variations of the same tired excuses.
Israel keeps diagnosing a marketing failure because admitting a policy failure would require someone in power to reflect on what Israel has actually done
And that brings us to the most important line in the supplied material.
Israel keeps responding to falling support by spending more on hasbara. But what if the message isn't what failed?
So if public support collapses, blame TikTok.
Blame the universities.
Blame young people.
Blame journalists.
Blame antisemitism.
Blame the algorithm.
Blame chatbots for not becoming enthusiastic enough spokesmachines.
Blame anything except the policies those people are reacting to.
Netanyahu calls this an eighth front because the language lets him keep pretending that Israel needs a better weapon and of course only he can protect Israel, it’s always been his schtick.
A bigger campaign. A cleverer contractor. More websites. Better search rankings. A friendlier chatbot answer.
But propaganda cannot permanently fix the political damage caused by the thing it is trying to sell. As the old adage goes, you can polish a turd, but you cannot make it shiny.
It can flood feeds. It can buy impressions. It can create sites, tune keywords, and try to influence which sources machines retrieve information from.
But it cannot make the public forget what has already been seen. It cannot convince the public, having seen what they have seen, to ignore the truth of their own eyes and ears.
The more obvious the attempts at this become, the less convincing the output will be.
Every campaign site identified damages the appearance of independence the network was meant to create. Every disclosed contract prompts another question about who is paying for it. Every investigation turns a supposed credibility operation into evidence of something completely manufactured instead.
We did.
We checked who built the sites.
We checked who paid the contractor.
We checked what the filed work plan actually said.
And we checked the limits of what the investigation proves.
The result is not that Israel secretly owns ChatGPT. The result is more embarrassing than that.
A state struggling to keep its credibility paid for a campaign meant to place its preferred framing inside the information environment people increasingly look to for answers. Then its own propaganda invited the public to trace the sources backwards. And people did.
That trail led through the websites, through Clock Tower X, and back to the State of Israel.
So Netanyahu's eighth front hasn't repaired Israel's credibility. It has produced a public demonstration of just how desperately his government is trying to manufacture it, because God forbid people see Israel for what it really is.
Nice try, Bibi.
For more on the Netanyahu instigated messes kicking off across all those war fronts right now and how so many of them are imploding back on him now, do stick with this channel for more, we do love to give him a roasting around here, such as this story right here.
Please do also hit like, share and subscribe if you haven’t done so already so as to ensure you don’t miss out on all new daily content, as well as spreading the word and helping to support the channel at the same time, which is very much appreciated, holding power to account for ordinary working-class people, and I will hopefully catch you on the next one. Cheers folks.