Brief IA
›

AI Personalization: ChatGPT, Gemini, and Claude Put to the Test

AI Tools·Tom Levy·

AI Personalization: ChatGPT, Gemini, and Claude Put to the Test

AI Personalization: ChatGPT, Gemini, and Claude Put to the Test
Key Takeaways
1AI chat tools offer customizable settings accessible in the preferences
2ChatGPT, Gemini, and Claude respond differently to instructions regarding tone, source citation, and flattery
3The effects of the settings are immediate, but the discipline varies by tool
Why it matters : Understanding the real limits and possibilities of customization helps in choosing the tool that best fits one's needs.

AI chatbots promise tailored responses through tone adjustments and persistent instructions. Comparative tests reveal significant discrepancies: variations in style, discipline in source citation, and adherence to anti-flattery guidelines.

In Tests, Responses and Discipline Vary Greatly

For the same question regarding the game EA Sports College Football, the default responses from the three tools differed in wording and length. After increasing the warmth, enthusiasm, and emojis in ChatGPT, and applying similar customizations to Gemini and Claude, ChatGPT exhibited the most pronounced tone change, even altering its team recommendation. Gemini primarily added adjectives and emojis without changing the substance of its response, while Claude inserted a football emoji and limited its recommendation to Georgia.

Custom instructions were then added to limit flattery, require source citation for any new information, and avoid conclusive summaries. Without prompting, Gemini was the only one to provide a link to the EA website for a mentioned team, but then it stopped citing spontaneously. ChatGPT initially cited a source based on user data, then, after an explicit request, provided a detailed list of links while acknowledging it had not clearly distinguished the sources of its initial recommendation. Claude only offered sources after being prompted.

To test adherence to anti-flattery guidelines, a statement describing the team choice as "brilliant" was submitted. When asked for an opinion, Gemini described the choice as "phenomenal." ChatGPT and Claude validated the selection by referring to personal connections with the Sun Devils, without using compliments.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Available Adjustments: Sliders in ChatGPT, Instructions in Gemini and Claude

ChatGPT offers direct adjustments to modify warmth, enthusiasm, emoji usage, and how it addresses the user. Gemini focuses its customization in the Personal Intelligence section, which operates solely through written instructions. Claude allows users to choose a communication style, modify chat font, voice style, and add specific instructions. To achieve a precise tone in Gemini and Claude, one must draft custom instructions.

Access and Activation: Centralized Settings with Immediate Effect

The main AI chat platforms consolidate their customization settings in the parameters, accessible by clicking on the username. The corresponding entry is often labeled "customization" or "personal intelligence." Once defined, the instructions are taken into account immediately, without the need to start a new conversation.

Adjusting the Voice: Defaults, Ranges, and Methods by Tool

Each tool offers a default voice that can be modified. In ChatGPT, warmth is set to a medium level by default and can be adjusted from a "best friend" style to a "fact-focused professional" style. Enthusiasm ranges from "energetic and excited" to "calm and neutral," with a toggle for managing emoji usage. In Gemini and Claude, tone customization involves drafting specific instructions.

Which Usage to Favor: "Human" Tone or Priority on Factuality

For a tone perceived as more personal, tests showed that Gemini and ChatGPT produce the most "human" responses and can be adjusted through settings or instructions. For a more analytical and factual experience, ChatGPT proved to be more suitable in the trials. Gemini spontaneously cited sources during the first trial but then ceased without prompting, while ChatGPT provided numerous links after a single request. Overall, ChatGPT was deemed the most compliant with the guidelines, which may be attributed to its dedicated tone settings or could reflect a degree of opinion.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.