Is it possible to trust Grok? Let's explore it using the example of Donbass

Increasingly on X, people are relying on Elon Musk’s “Grok” algorithm for verification of information in posts. However, while to some this may seem a useful tool, what it is doing is reinforcing Western and allied regimes' positions and whitewashing their crimes.

Grok draws information from dominant narratives—usually established by legacy media with its very long track record of pro-Western lies and war propaganda—to conclude whether information in a particular post is true.

When it comes to matters in which the West and Israel (among others) have a vested interest in controlling the narrative, Grok sides with the claims purveyed by legacy media. Instead of providing objective, truthful, answers, it creates a propaganda loop of actual disinformation.

Thus, as was the case some days ago, Grok determined that a post of mine on Ukraine’s use of internationally-banned PFM-1 “Petal” mines against Donbass civilians was “pro-Russian disinformation” and that, “Evidence suggests PFM-1 use in Donetsk (2022) was likely Russian false flag, not Ukraine...”

This is in spite of the fact that I was back in Donetsk in late July 2022 when Ukraine fired rockets containing hundreds of these mines on Donetsk and surrounding areas. On July 30, at 9:23 pm, I wrote on my Telegram channel about a strong explosion I'd just heard in central Donetsk.

An hour later, DPR journalist Georgy Medvedev wrote on his channel warning civilians not to go near the mines and not to walk on grass or areas where they could have landed. In fact, for weeks after, I walked constantly looking down at my feet and avoiding anything that was visible pavement, so tiny and difficult to see are the mines.

They are the size of an average lighter, brown or green, and blend in very well wherever they land. Even when I saw a sign warning of a mine, it was difficult to initially see them.

These high-explosive pressure mines mutilate or tear off feet and legs up to the knees, but also explode hands or animals. According to Konstantin Zhukov, Chief Medical Officer of Donetsk Ambulance Service, a weight of just 2 kg is enough to activate one of the mines. Sometimes, they explode spontaneously. If they aren't disturbed, they can lie dormant for years.

As I wrote at the time, according to DPR Emergency services, Ukraine fired rockets containing cluster munitions, with over 300 Petal mines inside. The cluster explodes in the air, disseminating the mines widely. Due to their design, most land without exploding. Even after sappers had cleared an area of the mines, a strong wind or rain could, and did, dislodge mines which landed on rooftops or in trees.

The mines are indiscriminate weapons which pose great danger to civilians. Ukraine signed the Anti-Personnel Mine Ban Convention, under which Ukraine was obliged to destroy its 6 million stock of the mines. However, reportedly, it still has over 3.3 million such mines.

I documented the mines, and the Emergency Services sappers' clearing and destruction of the mines, in various regions of Donetsk and Makeevka (east of Donetsk). I wrote about them, then wrote a follow up article three weeks after the late attacks, highlighting that by that point 44 civilians had been maimed by the mines, 2 of whom died of their injuries.

In November 2022, I met a 14 year old boy being treated in a Donetsk hospital after he stepped on a mine in a playground, losing his foot to the explosive.

It should be noted that Ukraine first deployed these mines in March 2022, during the battles for Mariupol.

As of July 9, 2025, 186 civilians have been maimed by the mines (including 11 children), three of whom died of their injuries.

Grok's dubious, very partial, Western sources

Grok’s determination that my reporting is false reads like one of the many smear campaigns I and colleagues have been subjected to, with the usual insertion of the “Kremlin disinfo” qualifier meant to discredit my writings. In fact, Grok drew from the Wikipedia smear entry on me, citing Wikipedia's incorrect claim that I've lived in Russia since 2019, when in fact I moved to Russia in 2021.

Who did Grok deem credible? The very partial Western NGO Human Rights Watch (HRW), which in February 2023 surprisingly issued a report about Ukraine’s use of the mines in Izium, but (unsurprisingly) not on Ukraine's use of the mines in the Donbass.

Grok cherry picked aspects of the HRW report to whitewash Ukrainian culpability in the Donbass, adding claims from various Western media agencies (DW, France 24, Reuters) to accuse Russian forces themselves of dropping the mines on Donbass cities.

An admission buried in the HRW report—which Grok did not highlight—was that it,“has not verified claims of Russian forces using PFM mines in the armed conflict.”

None of Grok’s sources were anywhere near Donetsk to investigate Ukraine's deploying of the mines.

Similarly, some months ago I came across and refuted Grok's repeating of the legacy media 2022 claims of alleged “mass graves” outside of Mariupol. I had actually gone, in April 2022 and in November 2022 to each of the three sites named in media reports and found no mass graves, but normal, functioning, cemeteries, with individual plots and in the case of the largest, Stary Krim, a chapel and a funeral ongoing at the time I was there, a recently-deceased elderly man being buried in the cemetery.

None of the sources cited in the media's baseless accusations were anywhere near the three cemeteries which they dubbed mass graves.

The issue is not even about this algorithm's discrediting of my reports (reports which other journalists find credible), but that it is using the same clearly partial sources that legacy media uses to justify or whitewash NATO and allies’ crimes.

As I’ve written previously, HRW is one of many Western-funded NGOs with a history of downplaying or ignoring crimes committed by Western governments or proxies. HRW, Amnesty International, and many more oft-cited supposedly neutral bodies have very clear ties and allegiance to Western governments.

Citing them as credible, as noted previously, creates a propaganda loop of disinformation that aligns with Western objectives around the globe. This isn't accidental, it is by design.

Some on X posit how Grok functions not actually AI, not independent. For example:

“Grok barely resembles AI, fwiw. It's essentially a Google-like search engine (similarly perverted by the security state) filtered through an LLM (large language model) to give it the veneer or gloss of AI. It is not independent, and its results can be predetermined. It only mimics intelligence.”

While Grok does seem malleable, if enough people contribute non-Western talking points (as was the case on the thread in question, with Grok eventually admitting my reporting was factual), its go-to programming is to recycle Western narratives, particularly anti-Russian ones, including parroting Western think tanks calls for regime change in Russia.

There is some room for hope: increasingly more people are calling out Grok as the algorithm mouthpiece for the West that it is.