A new study compared leading AI language models for B2B marketing and sales content. Claude narrowly outperformed Gemini, while ChatGPT lagged behind. All models struggled with text clarity.
Content teams and digital publishers relying on AI for B2B marketing and sales copy may want to reconsider their workflows after new research revealed persistent weaknesses in text clarity across top language models. Wortliga, commissioned by Sistrix, evaluated Google, OpenAI, and Anthropic models using 2,112 real-world B2B texts from marketing and sales.
The study assessed eleven content types, including social media posts, marketing emails, SEO blog articles, case studies, and whitepapers. Researchers scored outputs on readability, sentence structure, passive voice, and style. Claude Opus 4.7 achieved the highest average Wortliga-Score at 47.7 points, just ahead of Gemini 3.1 Pro (preview) at 46.8. GPT 5.5 trailed with 37.7 points. Wortliga considers texts only above 60 points to be truly clear.
According to study lead Gidon Wagner, the findings suggest that using AI-generated content without oversight in marketing and sales carries risks. The team used API access, neutral temperature settings, and eight standardized prompts to ensure fair comparisons. They observed that GPT 5.5, in particular, defaulted to formal and complex language unless given precise instructions, reducing overall clarity.
The research also highlighted the significant impact of prompt design on output quality. The models tended to mirror the tone and structure of the prompts they received. When prompts were formal or vague, the resulting texts became less understandable and sometimes failed to match the intended content type.
Sistrix founder Johannes Beus emphasized that clear language is now a critical factor for online visibility, not just a stylistic preference. He noted that both accessibility and machine readability depend on well-structured, understandable text.
Wortliga's data indicated that none of the tested AI models could consistently produce convincing B2B content without human intervention. Beus described the study as a candid snapshot of current AI capabilities and the ongoing need for editorial oversight.
These findings echo broader industry concerns about automated content quality. For example, recent experiments in digital publishing have explored new frameworks for measuring editorial success, as seen in Financial Express Digital's shift away from page views as a primary metric.
The final ranking for B2B marketing and sales text generation placed Claude Opus 4.7 first, followed by Gemini 3.1 Pro (preview), with GPT 5.5 in third.
