Here’s the prompt I’m going to start with:
Please list 5 long articles from distinguished newspaper publications and 5 articles from academic/scholarly journals (with DOIs - please give their DOI links if you can find them) on how and why European countries have been divesting from reliance upon American Big Tech companies like Microsoft, Amazon, Meta, Apple, Google since the 2000s. Please list these citations in APA format and please flag any citations that were difficult to format so I can review them. Thank you!
I’m going to try out https://scite.ai/
First, need to explore scite.ai’s ‘Assistant Settings’ where there’s a ‘Table Mode’ that gives you references in a table vs. a conversational format. Since I’m looking for straight references, I’m going to use Table Mode. There’s additional settings that are fantastic: you have a choice to limit sources older than the year 2000, you can say you’d like a long reply (500 words or more), you can specify if everything needs a reference or not (I’m going to “Let the assistant decide” because I feel like my prompt explicitly took care of this parameter).
Okay so everybody gets 3 free prompts per day.
Okay here we go:

Can you confirm the existence of the provided sources?
I’m going to scrap this and re-try a new prompt for scite.ai.
Are European countries divesting from reliance upon American Big Tech companies like Microsoft, Amazon, Meta, Apple, Google? Please cite 10 sources from academic journals or major news publications.
In the Assistant Settings in the field “Specify Reference Requirement,” I’ve set it to “Always Use References.”
Let’s see if that setting stops scite.ai from hallucinating references vs. my explicit prompt.
Scite.ai gave me one reference:
A day later, I’m going to take my second, more simple prompt with scite.ai’s default assistant settings to see what happens.
First, I really appreciate scite.ai took some time to build out an answer for me, and then extra time where scite.ai was informing me it was “verifying sources.” Next, reading through the answers, things seem pretty accurate (I follow this topic a lot), and there’s a link to one source:
Unfortunately, the rest of the answer in scite.ai is deliberately blurred: the citations are hidden behind a paywall.
From here, there are two avenues I might go: 1) see if signing up with an email to have a ‘basic’ tier with scite.ai will unlock the citations, or 2) try a different AI tool. For the purposes of sticking to one AI tool to explore, I’m going to sign up with my illinois.edu account (and potentially see if we have any connection/subscription with scite.ai yet - apparently, Stanford University does).
Upon signing up, you’re taken to a full page that feels like you must pay for the premium plan - 7 days free trial, $15 per month after that, enter your card information. Only when you move the mouse over the big blue logo with a tiny difficult-to-spot light-gray-colored arrow icon that typically means ‘back’ (see below) do you see a “back” button that lets you into your account dashboard with the ‘basic’ (not ‘free’) account tier.
Turns into this only upon hover.

This is considered a ‘dark pattern’ or ‘deceptive pattern’ - deceptive technology design that manipulates or tricks people to give their information or money. I could barely spot that little back arrow next to the scite.ai logo, and I have standard vision. And even if it were seen, most of my patrons would not think to explore that area of the page with their mouse to find the hover ‘back’ text that appears. AND EVEN if they did, the meaning of “Back” may - quite reasonably! - be perceived as a failed account creation; not a “Back to your Basic Account plan you’ve just created.” Web designers know exactly what they’re doing when they do stuff like this too; it’s contemptible. Might send a tip to the Dark Patterns Tipline (which is run by Stanford Digital Civil Society Lab?? Stanford University’s like the spider-man pointing at itself meme right now).
Moving on from unsolicited analysis of how accessible and/or exploitive scite.ai is once you start looking into their pricing & payment designs… Let’s see what citations I may have unlocked with the ‘Basic’ tier.
Again, heading into scite.ai with the question with no changes to the search settings: Are European countries divesting from reliance upon American Big Tech companies like Microsoft, Amazon, Meta, Apple, Google? Please cite 10 sources from academic journals or major news publications.
I did not unlock any additional citations, unfortunately. But I still got to unlock one, which was:
This citation is definitely relevant to my question. Its abstract is addressing European concerns over a lack of antitrust enforcement in the American tech industry, and adds a fascinating addition: German neoliberalism in the 21st century.
So far, all of scite.ai’s legitimate sources have been ‘current.’ This topic has been building steam during the 2010’s and especially after COVID, I would expect far more scholarly content about it. Particularly as a result of recent news showing significant European governmental and academic institutions divesting from American technology companies.
For my last question of the day (because I still only get 3 with a Basic Tier; what was the benefit of signing up at all? I cannot tell), instead of entering the same question and getting one last citation to analyze, I’m going to try that first long prompt again (with a couple edits), but this time with no changes to scite.ai’s default settings:
“Please list 10 articles from distinguished newspaper publications or academic/scholarly journals on how and why European countries have been divesting from reliance upon American Big Tech companies like Microsoft, Amazon, Meta, Apple, Google since the 2000s. Please list these citations in APA format and please flag any citations that were difficult to format so I can review them. Thank you!”
Let’s see if we can get some verified citations under a Basic Tier vs. the Free Tier I was in last time.
Oh my gosh! Success! scite.ai gave me an essay addressing the question and, when I scroll down under ‘Conclusion’, it actually gives me 10 citations. Let’s take a look at them.
As an aside, I don’t understand the completely false hallucinations in the beginning of this endeavor (unless these are also false, but scite.ai’s “verifying sources” status was moving as it was crafting its answer, so I’m more likely to believe these will be credible citations).
Here are ten articles in APA format, summarizing the literature on European countries divesting from American Big Tech:
Wow, it took a couple days to get here, and I’m so conflicted. Fiddling with Scite.ai’s Assistant Settings led to the creation of fake, hallucinatory citations. I thought my edits to the Assistant Settings made sense but even if they didn’t, scite.ai shouldn’t be able to hallucinate articles at all!
Rephrasing the prompt with Scite.ai’s Assistant Settings only gave me one citation, which was valid, but I asked for ten. Then, the next day, I tried again without fiddling with any Assistant Settings, and it put all but one citation beyond the ‘free tier’ of access that I was using. Knowledge has value, a central tenet of LIS, so I get it (sort of; I’m trying to access the citations, not the actual content!). Can I pay with my data (ie: creating an account) instead of actual money though? I created an account to get to the ‘basic tier’ to see if it would unlock my access to those citations.
While creating an account with scite.ai I flagged a vile dark/deceptive pattern that manipulates users into thinking they have to pay to complete the creation of their account.
With one last research question (both free and basic tiers only allow 3 research questions per day - what did I even sign up for?!), I rephrased my original, longer prompt, with no fiddling with the Assistant Settings, and finally got a list of 10 citations that I was able to analyze. They were all real, solid sources worth looking into regarding this research question, but it was painfuly obvious no one should use a citation found anywhere other than the source’s DOI site.