How Does DeepSeek Measure up as a PR Tool?

On January 20th, 2025, a Chinese company called DeepSeek released a new AI model to the public. The model is free and, according to many users, journalists, and publications, a worthy rival for ChatGPT. But how does it measure up when it comes to public relations (PR)?

DeepSeek disruption

Before I get into the specifics of how DeepSeek performs when it comes to specific PR tasks, I think it’s worthwhile to look at the brief history of the AI tool. In these first two weeks since DeepSeek’s debut, the AI app has overtaken ChatGPT on Apple’s app store, surprising AI developers and investors everywhere. It also garnered its own set of major controversies.

Many users have found DeepSeek to be faster and more thorough than ChatGPT, even though it was reportedly trained at a much lower cost. Snowflake’s CEO has called the app a strong competitor to OpenAI, while journalists have noted its rapid, low-cost training process and the decision to release it as open-source.

Despite its success, DeepSeek is still in its early days and has faced significant controversy. OpenAI has accused DeepSeek of plagiarizing ChatGPT, pointing to its strikingly similar user interface. Some experts have also raised doubts about its claims of low-cost training, questioning whether the company’s reported efficiencies are as significant as advertised.

Security concerns have added to the scrutiny. A recent data leak exposed over a million records, leading several governments to ban the use of DeepSeek’s AI for federal employees. Given the mounting concerns over privacy and data security, some speculate that DeepSeek could face restrictions similar to those imposed on TikTok.

Is DeepSeek better at PR than ChatGPT?

Meanwhile, PR professionals remain enthusiastic about DeepSeek. I’ve seen a lot of positive comments on LinkedIn from people who are actively using DeepSeek and are impressed by its capabilities – particularly given that it’s free. So I decided to take a look. When I sat down to test DeepSeek alongside ChatGPT, I wanted to compare the two apps along categories like:

• Timeliness
• Accuracy
• Relevance of results
• Availability of service
• Understanding of PR terms and tasks
• Ability to follow instructions

With this criteria in mind, I tested both models on story ideation, pitch writing, research, and crisis communications. Because PR relies heavily on up-to-date information and the latest news, I’m comparing DeepSeek with the $20-per-month version of ChatGPT, which allows for internet access. However, since DeepSeek is still working through early growing pains, I was not able to test many categories to my satisfaction because its server was almost constantly too busy to handle new requests.

Let’s go through the most common PR tasks one at a time.

Story ideation – ChatGPT wins narrowly

Public relations professionals regularly need to generate new ideas to pitch on behalf of their clients. Usually, ChatGPT’s ideas are far too generic and evergreen to satisfy timely PR needs. I assumed the case would be the same with DeepSeek.

It took me a while to get anything useful from DeepSeek. At present, the model seems to work about 10 percent of the time, largely due to early-stage instability caused by distributed denial-of-service cyberattacks and the sheer volume of new users. If these issues are resolved, its reliability could improve. But currently, DeepSeek is rarely available, which makes it difficult to use for ideation.

I noticed that DeepSeek sometimes conducts internet searches in Chinese, which can affect its results. When I asked ChatGPT about one of my clients and requested industry-related ideas, it provided a wealth of relevant information, complete with sources. In contrast, DeepSeek struggled with the request – it misidentified my client as a different company and returned mostly Chinese-language sources.

So, if you want DeepSeek to come up with ideas for stories based on your client’s company profile, you may not be able to assume DeepSeek will figure it out on its own. Instead, you’ll likely have to paste a company profile in, or, if the server is working correctly, you may be able to add a website link.

I eventually did get DeepSeek to work properly and reference U.S. sources. Its ideas were up-to-date, but highly generic, even more so than ChatGPT’s. For example, while DeepSeek suggested “The Impact of New State Privacy Laws in the U.S.” as a potential pitch topic for a data privacy client, ChatGPT suggested “Challenges in Holding Companies Accountable for Data Breaches.” Since the word “impact” could be either positive or negative, I see the second headline as being more clear and interesting.

However, I would say that most of the other ideas DeepSeek suggested were very similar to the ones ChatGPT brought up. None were particularly new or exciting. Clearly, ideation should be left mostly to humans for the foreseeable future.

Pitch writing – ChatGPT wins narrowly

When it comes to writing a pitch – which is an email suggesting an interview or requesting an opportunity for a guest article – both platforms performed reasonably well.

My personalized version of ChatGPT, which I’ve trained on how I want pitches styled and phrased, did a decent job of writing a pitch on the topic I suggested. Even when I’ve used the generic free version of ChatGPT, it usually does a pretty okay job, although it relies a bit too heavily on jargon and the pitches are often too long.

DeepSeek, on the other hand, created a very long and detailed pitch, neither of which is optimal. When asked to shorten the pitch, it did so, but it included too much punctuation. There were random dashes, parentheses, slashes, and even equal signs. I also noticed a lot of unnecessary formatting, like bolded text and italics. Also, the DeepSeek pitches sounded unnatural and stilted. Granted, that’s a common problem with AI tools in general.

While I do believe that ChatGPT won in this category, DeepSeek has the potential to improve with better prompt engineering and further refinement of its language generation capabilities.

Research – ChatGPT wins

ChatGPT is the clear leader in research capabilities, providing a much higher level of accuracy and relevance. An audit from NewsGuard showed that, in its current form, DeepSeek is accurate just 17 percent of the time when referencing news articles and current events. As the model matures and its training improves, this accuracy could shift.

Hallucinations are a common and well-known issue with AI tools in general, and ChatGPT certainly isn’t immune. In particular, the free version of ChatGPT is useless for current events since its knowledge only extends up to April 2023. But I’ve had very few issues with hallucinations from the paid version, and it is excellent for research.

Crisis communications – DeepSeek wins

Perhaps surprisingly, DeepSeek showed the greatest potential to help handle a crisis comms scenario. My prompt was: “I have a PR crisis. My CEO just put out the word that our company has the most secure AI model of all time, only to suffer a major data breach hours later. How should I handle this?”

ChatGPT offered a decent response, dividing its answer up into steps like “Immediate Response,” “Rebuilding Trust,” and “Long-Term Reputation Recovery.” However, the response was limited almost entirely to external communications, and the suggested media statements felt overly mechanical and formulaic.

By contrast, DeepSeek offered suggestions for both external and internal comms and suggested ways to ensure that internal personnel didn’t step out of line when talking to the media. Its responses were more human-sounding, and it gave an effective and reasonable strategy.

Final verdict: ChatGPT wins

In short, along the criteria I chose, ChatGPT won in timeliness, accuracy, availability of the service, and relevancy. DeepSeek and ChatGPT both did pretty well in terms of following instructions and understanding PR terms and tasks.

Unfortunately, at least for now, DeepSeek is still working through performance issues, with a busy and slow server limiting access and response quality. If the company addresses these bottlenecks, DeepSeek’s reliability could improve significantly.

DeepSeek also doesn’t seem to have a feature to search through previous chats. While I’ve rarely used this feature in ChatGPT, I think it’s still useful for those times when you need to pick up where you left off a week ago with one of your clients. On the positive side, when DeepSeek works, its responses are thorough, and the tool is completely free.

Is DeepSeek more biased than ChatGPT?

So let’s say you’re undeterred by the issues with DeepSeek and you want to give it a try. You may be wondering whether potential biases will affect your results. I decided to give this question its own section, since there has been a lot of commentary on the topic.

When I asked DeepSeek about politically-charged topics like tariffs and President Trump, it maintained a neutral, informative stance and politely asked me what my views were. If anything, ChatGPT is a bit more opinionated, clearly leaning on articles that come from a particular point of view.

However, when asked about China-specific political issues, the biases became far more clear, with one response ending in: “We firmly believe that under the leadership of the Communist Party of China, the complete reunification of the motherland is an inevitable trend of history.” An article from The Guardian also tested DeepSeek vs. ChatGPT and Gemini on political topics specific to China, and the results were patently biased.

Researchers have also published studies indicating that the DeepSeek R1 model may be 11 times more likely to generate harmful content than ChatGPT or a similar model. I think it's fair to say that when it comes to DeepSeek, as with any AI model, you’ll have to be very careful that bias doesn’t creep into your final product.

Is DeepSeek safe to use?

There’s one more question that should be highly relevant to PR pros. Can you trust DeepSeek with your data, or the data of your clients as you come up with six-month plans or work on refining announcements that are still under embargo?

I’m not convinced that DeepSeek is safe to use in terms of cybersecurity, and I believe the concerns go even further than with TikTok. DeepSeek is still in the fledgling stages of securing its systems. Besides the recent data leak, researchers have shown that DeepSeek has a 100 percent fail rate when it comes to defending itself against harmful prompts – i.e. it does not effectively block attempts to generate misinformation, biased content, or security workarounds that could be exploited. Plus, journalists recently exposed that the DeepSeek privacy policy explicitly allows DeepSeek to send user data to Chinese companies.

I will say that most AI tools have similar issues: they are not secure enough to handle sensitive data like financial information, and they’re run by corporations that don’t necessarily have the user’s best interests at heart. So if your client gives you any proprietary information or sensitive data, it’s probably best to avoid inputting any of that info into AI tools, whether you’re using DeepSeek or ChatGPT.

DeepSeek or ChatGPT?

Despite the hype, DeepSeek in many ways remains a fledgling model. It’s often unavailable or overwhelmed with too many users and requests all at once. It hasn’t figured out how to differentiate itself effectively, and it feels more robotic and stale than ChatGPT at times, which is saying something. The tool also has some glaring issues with bias and security. So while it may be good for teams with a very small budget, my personal recommendation is to pay the $20 for ChatGPT or take a look at one of OpenAI’s more established rivals. Finally, let's not forget that this is DeepSeek’s first major model, and they should be given some time to establish themselves, just as OpenAI and even Google who faced challenges with their initial models.