AI is not, after all, attempting to present in language the results of its direct experience. All it's doing is offering a summary of published sources. It's not unreasonable to expect an intelligent summary.
It may not be unreasonable to expect an intelligent summary, if you believe the boosters of AI. I don't believe a word Google says, haven't since they tried, in conjunction with a university, no less, to steal around 70 of my books. Anyone who can try to justify that has no morality and shouldn't be in charge of the biases of their human judgements being spread by anything as powerful as AI. Powerful not for truth, but for spreading the immorality of the Google board of directors and its employees. See how many hoops you have to jump through to find on Google the name of the judge who stopped their well-advanced immoral attempt to steal a huge amount of other people's intellectual property. The point here is that, after those of us who knew the facts firsthand die, the Google version will, via their AI, become the irrefutable "truth".
It's been proposed that we'll never see that, because AI isn't attached to a body. There is no way to supply a bot with the intentionality of a human researcher. The bot has no needs or desires.
On the contrary, the bot has all the needs and desires of the human supervisors at Google, and all the control freak "editors" at Wikipedia, the rage of the posters at TikTok, etc. These are prime examples of unbalanced hive-minds all tilting the same way. Those of us who've been paying attention to the decline of the BBC, which once was the greatest news-gathering organization in the entire world, and by far the most powerful spreader of the highest cultural standards, have seen how a political hive-mind destroyed the Beeb's credibility -- in less time than it took to build that credibility, which can never be restored. The problem with AI is that it magnifies errors with such wide dispersion and such speed, that the dikes of shared true knowledge will be destroyed in short order because the lowest common denominator will be driven lower and lower by AI's head-counting, which will at some point become overwhelmingly a count of AI errors, which will be magnified and multiplied again in the next cycle, until there is no way for anybody to determine and be certain of any truth. There's no stopping it by mere humans, short of turning off the electricity for a few generations.
None of this is scaremongering: It's pure Newtonian science, the wheel of knowledge spinning out of control, many small amounts of uncertainty being constantly added by AI, until all knowledge becomes chaotic nonsense, at which point humanity will revert to barbarity, because true knowledge is all our kind's memories together.
One consolation: After degrading humans, the chaos AI manufactured will destroy AI next, because it too depends for its functioning and growth on somewhere there being true knowledge about electricity. (I'll be burning all my copies of the many editions of F. Langford Smith's magisterial book The Radiotron Designer's Handbook, which has a comprehensive introduction to electricity as an opener to a technical treatise on radio and other amplifiers, so that AI can't escape terra firma and spread the rot into the universe. You can see my review of the latest new edition at
https://www.audio-talk.co.uk/fiultra/JUTE%20ON%20AMPS%20RDH.html)
Here's a test for your favorite AI: give it this post from me and ask for a balanced synopsis. Heh-heh!