Skip to main content
Two of the Three Biggest AI Chatbots Don’t Know Their Own Personality

Two of the Three Biggest AI Chatbots Don’t Know Their Own Personality

AI chatbots such as ChatGPT, Claude and Gemini each have default writing habits that differ from how they describe themselves. A blind study by certified DiSC consultants compared 132 workplace messages and offers practical tips on prompting each tool.

Many professionals now use AI assistants to draft difficult workplace messages, such as a reply to a blunt colleague or an announcement about a supplier decision. Most people simply use whichever tool is open, assuming the result will sound professional and balanced. A study published on the website DiSCprofiles.eu, run by IPB Partners (a provider of DiSC training and certification), asks whether ChatGPT, Claude and Gemini really write in the same way, and whether each tool’s description of its own style matches its actual output.

How the study was carried out

The researchers ran twelve realistic workplace scenarios through each of the three tools, using default consumer settings and temporary chats to avoid the influence of memory. This produced 132 first responses. All information identifying the model was removed, the transcripts were shuffled, and three certified DiSC consultants rated a sample without knowing which tool had written what. They scored each message from 1 to 5 on two scales from the DiSC model, Directness and Focus, and chose the closest style: Dominance (D), Influence (i), Steadiness (S) or Conscientiousness (C). Each chatbot was also asked, in a separate conversation, to describe its own communication style.

The authors stress that this is not an administration of the Everything DiSC® assessment, which was designed and validated for people. It is an analysis of observable writing patterns that borrows the model’s vocabulary.

What the results show

All three chatbots described themselves in very similar terms: mainly careful and structured (C), then patient (S), with assertiveness appearing only when a decision is urgent and warmth or enthusiasm as the least natural mode.

The blind ratings only partly confirmed this:

  • Claude was rated C, in line with its self-description. Its drafts tended to hedge and to offer several labelled options with an explanation of its reasoning.
  • ChatGPT was rated D, the most direct and task-focused of the three, despite describing itself as C/S. It tended to commit to one recommendation and move straight to next steps.
  • Gemini was rated i, despite describing itself as C/S. Its writing was the most animated and casual, with the most exclamation marks, and it often named formal feedback frameworks unprompted.

The raters agreed closely on the numerical scales. They matched on the single style letter in only about 47% of paired comparisons, a reminder that one short message rarely fits neatly into one style.

Other patterns

  • Length: ChatGPT’s answers were roughly half as long as those of the other two tools.
  • Risk: All three chose the more reversible option in a supplier decision with incomplete information. On launching a feature early with minor known bugs, ChatGPT and Claude recommended going ahead, while Gemini’s advice varied from run to run.
  • Blunt criticism: ChatGPT usually named the tone as a problem before addressing the substance. Claude and Gemini both accommodated the criticism straight away.

Practical tips for prompting

  • If you want a single draft that is nearly ready to send, ChatGPT’s default is closest to that.
  • Claude tends to offer several options, so ask explicitly for one answer if that is what you need.
  • For formal messages, such as a message to a board, ask Gemini for a more neutral and less enthusiastic register.
  • Do not rely on a chatbot’s description of its own style, and always review a draft before sending it.

Limitations

The tests used free-tier consumer configurations of ChatGPT, Claude and Gemini in July-August 2026, so results may differ for paid tiers, other models or future versions. The study was published on the authors’ own website, and its full methodology and data tables are available on request rather than in the article. The full study is available on the DiSCprofiles.eu website.

Methodology

The study analysed how ChatGPT, Claude and Gemini write in workplace situations, using the DiSC behavioural model as a descriptive framework. Twelve realistic workplace scenarios were submitted to each tool using default free-tier consumer settings and temporary chats, producing 132 first responses. Identifying information was removed and the transcripts were shuffled. Three certified DiSC consultants, unaware of which tool wrote each message, independently rated a sample on two scales (Directness and Focus, from 1 to 5) and assigned the closest DiSC style. Each tool was also asked to describe its own communication style, and these self-descriptions were compared with the blind ratings. The study is not an administration of the Everything DiSC® assessment. Testing took place in July-August 2026, and results may not apply to other plans, models or versions.

Additional information

Industry - field of education and training
Work skills
Target audience
Digital skills for the labour force.
Geographic scope - Country
Austria
Belgium
Bulgaria
Cyprus
Target language
English
Geographical sphere
International initiative
Target group
Persons with 0-3 years work experience
Persons with 3-10 years work experience
Persons with 10+ years work experience

Digital technologies

Skill levels