On January twentieth, 2025, a Chinese language firm referred to as DeepSeek launched a brand new AI mannequin to the general public. The mannequin is free and, based on many customers, journalists, and publications, a worthy rival for ChatGPT. However how does it measure up in relation to public relations (PR)?
DeepSeek disruption
Earlier than I get into the specifics of how DeepSeek performs in relation to particular PR duties, I believe it’s worthwhile to have a look at the temporary historical past of the AI software. In these first two weeks since DeepSeek’s debut, the AI app has overtaken ChatGPT on Apple’s app retailer, stunning AI builders and traders all over the place. It additionally garnered its personal set of main controversies.
Many customers have discovered DeepSeek to be sooner and extra thorough than ChatGPT, despite the fact that it was reportedly educated at a a lot decrease value. Snowflake’s CEO has referred to as the app a robust competitor to OpenAI, whereas journalists have famous its fast, low-cost coaching course of and the choice to launch it as open-source.
Regardless of its success, DeepSeek continues to be in its early days and has confronted important controversy. OpenAI has accused DeepSeek of plagiarizing ChatGPT, pointing to its strikingly comparable person interface. Some specialists have additionally raised doubts about its claims of low-cost coaching, questioning whether or not the corporate’s reported efficiencies are as important as marketed.
Safety considerations have added to the scrutiny. A current information leak uncovered over one million data, main a number of governments to ban the use of DeepSeek’s AI for federal workers. Given the mounting considerations over privateness and information safety, some speculate that DeepSeek might face restrictions much like these imposed on TikTok.
Is DeepSeek higher at PR than ChatGPT?
In the meantime, PR professionals stay smitten by DeepSeek. I’ve seen lots of constructive feedback on LinkedIn from people who find themselves actively utilizing DeepSeek and are impressed by its capabilities – significantly on condition that it’s free. So I made a decision to have a look. After I sat down to check DeepSeek alongside ChatGPT, I needed to check the 2 apps alongside classes like:
- Timeliness
- Accuracy
- Relevance of outcomes
- Availability of service
- Understanding of PR phrases and duties
- Capacity to observe directions
With this standards in thoughts, I examined each fashions on story ideation, pitch writing, analysis, and disaster communications. As a result of PR depends closely on up-to-date info and the newest information, I’m evaluating DeepSeek with the $20-per-month model of ChatGPT, which permits for web entry. Nevertheless, since DeepSeek continues to be working by means of early rising pains, I used to be not capable of take a look at many classes to my satisfaction as a result of its server was nearly consistently too busy to deal with new requests.
Let’s undergo the commonest PR duties one after the other.
Story ideation – ChatGPT wins narrowly
Public relations professionals repeatedly have to generate new concepts to pitch on behalf of their shoppers. Often, ChatGPT’s concepts are far too generic and evergreen to fulfill well timed PR wants. I assumed the case could be the identical with DeepSeek.
It took me some time to get something helpful from DeepSeek. At current, the mannequin appears to work about 10 % of the time, largely as a consequence of early-stage instability attributable to distributed denial-of-service cyberattacks and the sheer quantity of latest customers. If these points are resolved, its reliability might enhance. However at present, DeepSeek is never accessible, which makes it tough to make use of for ideation.
I seen that DeepSeek typically conducts web searches in Chinese language, which might have an effect on its outcomes. After I requested ChatGPT about certainly one of my shoppers and requested industry-related concepts, it supplied a wealth of related info, full with sources. In distinction, DeepSeek struggled with the request – it misidentified my consumer as a distinct firm and returned largely Chinese language-language sources.
So, if you’d like DeepSeek to give you concepts for tales based mostly in your consumer’s firm profile, it’s possible you’ll not be capable to assume DeepSeek will determine it out by itself. As an alternative, you’ll seemingly have to stick an organization profile in, or, if the server is working accurately, you could possibly add a web site hyperlink.
I ultimately did get DeepSeek to work correctly and reference U.S. sources. Its concepts had been up-to-date, however extremely generic, much more so than ChatGPT’s. For instance, whereas DeepSeek urged “The Impression of New State Privateness Legal guidelines within the U.S.” as a possible pitch subject for an information privateness consumer, ChatGPT urged “Challenges in Holding Firms Accountable for Knowledge Breaches.” Because the phrase “impression” could possibly be both constructive or detrimental, I see the second headline as being extra clear and attention-grabbing.
Nevertheless, I might say that many of the different concepts DeepSeek urged had been similar to those ChatGPT introduced up. None had been significantly new or thrilling. Clearly, ideation ought to be left largely to people for the foreseeable future.
Pitch writing – ChatGPT wins narrowly
On the subject of writing a pitch – which is an e mail suggesting an interview or requesting a possibility for a visitor article – each platforms carried out moderately nicely.
My personalised model of ChatGPT, which I’ve educated on how I need pitches styled and phrased, did a good job of writing a pitch on the subject I urged. Even after I’ve used the generic free model of ChatGPT, it often does a reasonably okay job, though it depends a bit too closely on jargon and the pitches are sometimes too lengthy.
DeepSeek, alternatively, created a really lengthy and detailed pitch, neither of which is perfect. When requested to shorten the pitch, it did so, however it included an excessive amount of punctuation. There have been random dashes, parentheses, slashes, and even equal indicators. I additionally seen lots of pointless formatting, like bolded textual content and italics. Additionally, the DeepSeek pitches sounded unnatural and stilted. Granted, that’s a standard drawback with AI instruments usually.
Whereas I do imagine that ChatGPT received on this class, DeepSeek has the potential to enhance with higher immediate engineering and additional refinement of its language technology capabilities.
Analysis – ChatGPT wins
ChatGPT is the clear chief in analysis capabilities, offering a a lot greater stage of accuracy and relevance. An audit from NewsGuard confirmed that, in its present type, DeepSeek is correct simply 17 % of the time when referencing information articles and present occasions. Because the mannequin matures and its coaching improves, this accuracy might shift.
Hallucinations are a standard and well-known concern with AI instruments usually, and ChatGPT actually isn’t immune. Particularly, the free model of ChatGPT is ineffective for present occasions since its data solely extends as much as April 2023. However I’ve had only a few points with hallucinations from the paid model, and it’s wonderful for analysis.
Disaster communications – DeepSeek wins
Maybe surprisingly, DeepSeek confirmed the best potential to assist deal with a disaster comms situation. My immediate was: “I’ve a PR disaster. My CEO simply put out the phrase that our firm has essentially the most safe AI mannequin of all time, solely to endure a serious information breach hours later. How ought to I deal with this?”
ChatGPT supplied a good response, dividing its reply up into steps like “Instant Response,” “Rebuilding Belief,” and “Lengthy-Time period Fame Restoration.” Nevertheless, the response was restricted nearly totally to exterior communications, and the urged media statements felt overly mechanical and formulaic.
In contrast, DeepSeek supplied recommendations for each exterior and inside comms and urged methods to make sure that inside personnel didn’t step out of line when speaking to the media. Its responses had been extra human-sounding, and it gave an efficient and cheap technique.
Last verdict: ChatGPT wins
In brief, alongside the standards I selected, ChatGPT received in timeliness, accuracy, availability of the service, and relevancy. DeepSeek and ChatGPT each did fairly nicely by way of following directions and understanding PR phrases and duties.
Sadly, no less than for now, DeepSeek continues to be working by means of efficiency points, with a busy and gradual server limiting entry and response high quality. If the corporate addresses these bottlenecks, DeepSeek’s reliability might enhance considerably.
DeepSeek additionally doesn’t appear to have a function to look by means of earlier chats. Whereas I’ve not often used this function in ChatGPT, I believe it’s nonetheless helpful for these instances when it’s good to decide up the place you left off every week in the past with certainly one of your shoppers. On the constructive aspect, when DeepSeek works, its responses are thorough, and the software is totally free.
Is DeepSeek extra biased than ChatGPT?
So let’s say you’re undeterred by the problems with DeepSeek and also you need to give it a strive. Chances are you’ll be questioning whether or not potential biases will have an effect on your outcomes. I made a decision to present this query its personal part, since there was lots of commentary on the subject.
After I requested DeepSeek about politically-charged subjects like tariffs and President Trump, it maintained a impartial, informative stance and politely requested me what my views had been. If something, ChatGPT is a little more opinionated, clearly leaning on articles that come from a selected perspective.
Nevertheless, when requested about China-specific political points, the biases grew to become way more clear, with one response ending in: “We firmly imagine that beneath the management of the Communist Social gathering of China, the whole reunification of the motherland is an inevitable pattern of historical past.” An article from The Guardian additionally examined DeepSeek vs. ChatGPT and Gemini on political subjects particular to China, and the outcomes had been patently biased.
Researchers have additionally revealed research indicating that the DeepSeek R1 mannequin could also be 11 instances extra seemingly to generate dangerous content material than ChatGPT or an analogous mannequin. I believe it is truthful to say that in relation to DeepSeek, as with every AI mannequin, you’ll need to be very cautious that bias doesn’t creep into your closing product.
Is DeepSeek protected to make use of?
There’s yet another query that ought to be extremely related to PR execs. Are you able to belief DeepSeek along with your information, or the info of your shoppers as you give you six-month plans or work on refining bulletins which can be nonetheless beneath embargo?
I’m not satisfied that DeepSeek is protected to make use of by way of cybersecurity, and I imagine the considerations go even additional than with TikTok. DeepSeek continues to be within the fledgling levels of securing its techniques. Apart from the current information leak, researchers have proven that DeepSeek has a 100% fail price in relation to defending itself in opposition to dangerous prompts – i.e. it doesn’t successfully block makes an attempt to generate misinformation, biased content material, or safety workarounds that could possibly be exploited. Plus, journalists just lately uncovered that the DeepSeek privateness coverage explicitly permits DeepSeek to ship person information to Chinese language firms.
I’ll say that the majority AI instruments have comparable points: they don’t seem to be safe sufficient to deal with delicate information like monetary info, and so they’re run by companies that don’t essentially have the person’s finest pursuits at coronary heart. So in case your consumer provides you any proprietary info or delicate information, it’s in all probability finest to keep away from inputting any of that information into AI instruments, whether or not you’re utilizing DeepSeek or ChatGPT.
DeepSeek or ChatGPT?
Regardless of the hype, DeepSeek in some ways stays a fledgling mannequin. It’s typically unavailable or overwhelmed with too many customers and requests suddenly. It hasn’t found out how you can differentiate itself successfully, and it feels extra robotic and off than ChatGPT at instances, which is saying one thing. The software additionally has some obvious points with bias and safety. So whereas it might be good for groups with a really small finances, my private suggestion is to pay the $20 for ChatGPT or check out certainly one of OpenAI’s extra established rivals. Lastly, let’s not overlook that that is DeepSeek’s first main mannequin, and they need to be given a while to ascertain themselves, simply as OpenAI and even Google who confronted challenges with their preliminary fashions.