By now, you’ve probably seen the flood of “artificial intelligence (AI)-powered research assistants” promising to do the reading, synthesis, and maybe even the strategic thinking for you. Tools like ChatGPT and Perplexity claim they can conduct “deep research,” leading some to ask if researchers will soon be replaced.
So naturally, someone in a government or corporate role might start to wonder: Are international affairs researchers next on the chopping block?
Short answer: No. You’re safe. For now.
But it’s not just about job security—it’s about understanding what international affairs professionals uniquely bring to the table, especially in objective research and cross-cultural communication. So, let’s look at how AI currently compares to your unique strengths.
What AI thinks Is “research” is…not what we do
AI assistants like ChatGPT and Perplexity (two of the more well-known among a growing field of tools) describe their “deep research” as something iterative, logical, and source-based. That sounds promising—until you actually try using them for the kinds of questions we handle every day.
Perplexity describes its deep research process as:
“Iteratively searches, reads documents, and reasons about what to do next, refining its research plan as it learns more…”
ChatGPT describes its deep research process as:
“Plan[s] and execute[s] a multi-step trajectory to find the data it needs…backtracking and reacting to real-time information where necessary.”
This sounds good. But in reality, these language models are English based and use the logic of a Western English speaker. Unfortunately, their iterative process is not de-biasing, but just the opposite.
In particular, AI isn’t built to question the assumptions behind what it’s finding. If the internet claims “no data is available,” AI shrugs and repeats it. Meanwhile, a trained researcher thinks: Really? Or is it just buried behind a different keyword on a provincial university site?
Even when you attempt to direct the AI tool to use only foreign language sources available in open internet archives like the Wayback Machine, most can’t or won’t search the archive effectively for tailored insights. They’re designed to summarize—not explore.
Even for the type of question Perplexity and ChatGPT claimed deep research excelled, the bias towards English language sources confused their findings. When asked to identify top foreign companies in a niche technology field using only foreign-language sources? Both ChatGPT and Perplexity defaulted to summarizing English-language reports.
After multiple nudges, they finally redid their searches and—surprise—returned a different top ten list when using foreign-language sources. The top three companies changed entirely, revealing something any international affairs professional knows: source selection is fundamental to representing regional perspectives.
Worse, when asked for a comprehensive list of foreign-language sources on a specific issue, neither ChatGPT nor Perplexity could deliver—despite repeated prompting. Instead of surfacing the full range of material, they kept trying to provide their version of the “best” answer. Helpful? Occasionally. But this is exactly where AI assistants turn from tools into filters—removing nuance, not revealing it.
Cross-linguistic, cross-cultural research still needs human-in-the-loop
One of the subtler but crucial skills we bring to answering questions about another country or international institution is how we conduct the search. Not just which keywords we use, but in which language, in which timeframe, on which platforms—and with which understanding of local context.
AI research assistants are still heavily biased toward English sources. Even when you enter non-English search terms, most tools struggle to crawl foreign-language websites effectively or understand the cultural logic behind those sites’ structures and discourse.
This is why we still manually search through Baidu, Naver, or Yandex—because the way a Russian researcher frames a biotech breakthrough isn’t how a U.S. think tank does. AI lacks the cognitive flexibility (for now) to spot that difference and pivot accordingly.
AI is getting better at customer service in multiple languages, but let’s not confuse tone-matching with cultural fluency. When asking AI about other cultures, it still reflects the bias of its training data.
AI can play out a cross-cultural negotiation scenario based on business language textbooks. This is certainly useful for language learners. But unless you are going to record all your negotiations to train the AI, there is nothing in the training model that matches the real thing.
In that regard, when you’re conducting an interview or negotiation abroad, you’re drawing on years of lived cultural exposure, linguistic subtlety, and emotional calibration. That’s not just a ”soft skill”—it’s a “hard to code skill.”
What about foreign market research? It still needs your data from the survey you designed. Even then, AI can’t tell you why participants answered the way they did.
Quick-Turn Questions? AI Fumbles
Another area where AI falls short is in the fast-paced, high-context environment of real-world international affairs.
Let’s say you’re asked to brief about a recent trend such as protests from an earthquake last month. AI might get you a few articles—from the last time its training data was updated. As of May 2025, unless you’re paying for a subscription, you may not get access to any internet postings after mid-2024. Talk about failing to be timely.
When using one of the limited free deep research queries, both Perplexity and ChatGPT could find information within the last year, but it still took multiple rounds of telling the AI to stop believing other English sources alleging the information was unavailable.
AI assistants might be partners—but not substitutes
The question isn’t whether we’ll use AI. It’s how. And when. IAThinker explained four steps to use specific AI assistants to answer questions about global trends in technology in our Analyst’s Roadmap. We showed you a detailed example of the method called Technical Expertise Builder in our video. These are examples of partnership, not substitutes.
Some large companies that have attempted to outsource tasks to contractors promising huge cost savings by using AI have had less that satisfactory results. Deloitte’s Global Outsourcing Survey from 2024 found that “Despite the high expectations from AI-powered outsourcing, the tangible benefits remain modest. Less than half of the organizations report productivity gains, and only 25% are seeing reduction in cost.”
A February 2025 Nature article about how researchers use AI found that of the 4,946 researchers surveyed worldwide—27% of whom were early-career professionals—they aren’t really using these tools much in their day-to-day work. Only 45% of the first wave of respondents (1,043 researchers) said that they had actually used AI to help with their research, and the most common uses they cited were translation, proofreading, and editing manuscripts.
As an important finding for international affairs professions—the Nature article found that there are clear differences across countries and disciplines, with researchers in China and Germany, as well as computer scientists, being the most likely to use AI in their work. How will this impact cross-cultural communication?
Looking Ahead: What Might Change?
One space worth watching is the development of foreign-language-specific AI tools like DeepSeek, that were probably largely trained with Chinese-language sites using local cultural logic. Concerns around censorship are real—but so is the potential. For example, asking DeepSeek to use only Weixin to rank the most influential Chinese voices on a specific technology may yield more accurate insights that are otherwise invisible to English-centric tools.
Shutdown
AI research assistants are improving, yes. They’re helpful, yes. But they’re still not built to do the kind of research international affairs professionals specialize in: objective, thorough, bilingual research that requires not just data, but discernment.
So don’t hang up your analyst hat just yet. In fact, the rise of AI might make your role more vital—most AI tool developers know nothing about what you do and have no way to build a tool to replace you.
Especially hard to replace are the colleagues among us who’ve scrolled through 20 pages of keyword variations in Uzbek at 2 a.m. to find the one interview clip that confirms the whole story.
You know who you are. We thank you.
