Case Study: Text Generation
Generate text. Save time. Improve SEO.
A major digital marketing player was looking for ways to innovate in a crowded Search Engine Optimisation (SEO) market. The competition demanded more impressive results from their SEO activities such as writing engaging titles and descriptions for client’s web pages, which numbered in the tens of thousands per client. Traditional approaches involved logging agency hours to create memorable and high-performing content, but the sheer scale of the website made the efficacy of this undertaking financially untenable.
Remi AI used generative adversarial networks (GANs) and Long Short Term Memory (LSTM) networks to produce engaging titles and descriptions for each of the ~9000 web pages for one pilot client’s site alone. A combination of tried and tested methods were coded into rules to accompany the generative AI models such as a title containing 3 parts: Title 1 | Title 2 | Title 3. The outputs were scored using relevancy metrics to rank the titles with the highest ranking accepted for use by the client.
Preliminary results were extremely promising: the AI approach can produce engaging titles for a large number of web pages, an activity that would previously have been done by a human. The outcome is such that the digital marketing client saved labour while maintaining a high quality, and their clients’ were able to receive more service for their money.