Developers, who are preparing the system for a dummy trial by Edexcel on GCSE-type scripts next year, believe it is accurate and economical. But authors, academics and teachers have already attacked the plans, fearing the system will fail to appreciate sophisticated essays and encourage simplified teaching.
The Pearson Knowledge Technologies system rates essays for content, syntax, style and spelling without human help. It is already used in several parts of the United States, where it has attracted controversy. John Mortimer, the author and playwright, called the concept "lunacy".
"How can a computer listen to the sound of words? How can it decide whether ideas are original or not?" he asked.
But Jerry Jarvis, managing director of Edexcel, said: "You start off sceptical but after a while you realise there's something in it. The makers believe it is more accurate than human beings."
While he conceded that Britain was not yet fully ready for e-marking of GCSE and A-level essays, he plans to hold the first-ever UK dummy trials of the software next year. Edexcel's human examiners already do more than 70 per cent of their marking on computers.
"I believe I'm doing the best for students. Others will disagree," Mr Jarvis said.
Geoff Barton, a fellow of the English Association, said the system might free up human markers to focus on finer qualities of writing, while the computers dealt with common spelling and grammar problems.
But Michael Morpurgo, the award-winning children's author, said the software was a symptom of "depersonalised" education.