Anyway, the return messages for the Google News rejects included issues like articles having too many bullet points, too many single sentence paragraphs, or simply being too short.

What concerns me is that here we have an automated system deciding what is news and what is not based on the structure of the article alone. No one is actually reading the article to decide if it's news.

It is only natural that when we are deciding whether a news source is reliable or not we base judgements on appearance. That's how we tell between a comic and a serious broadsheet (sorry, compact). We look at the style and font size perhaps. In broadcast terms we might make judgements based on accent and pace of delivery - is there a rock music backing track to the news or is the newsreader sat behind a desk wearing a tie?

But - and this is the crucial bit - we thankfully do recognise that a casually dressed, standing newsreader or a mobile phone display can still deliver valid news.

Here's another case of machines not knowing best:

I was searching on the web for a speech by an MP (one about immigration and asylum policy), but the server of the organisation whose computer I was using told me the web site was blocked because the content was 'deemed inappropriate'.

You can imagine there might be some racist or otherwise dodgy language in a speech about immigration, but not in this case. The on-screen message told me the site was blocked because the system thought it was a 'job search'.

With the help of the IT department I managed to get round the content blocker and view the site that had been 'deemed inappropriate' by the unhelpful little robot hiding in the PC. It was, as I expected, a perfectly safe and straightforward speech.

"There must be a certain pattern of words in the speech that made the system think it was a job search site," said the IT man. But he couldn't spot the offending word pattern.

So the software had decided I shouldn't be allowed access to this item.

Yet again, here was a decision being made by a piece of code looking at the structure of the article or the word patterns - but not actually reading or understanding the text.

Doesn't it make you wonder who's in charge here? Man or machine?

One of the great PR skills is being able to write material for different target audiences. More and more we also need to be thinking about how search engines will read our articles.

At Actuality Media we can claim some success at producing web copy that hopefully makes a good read and boosts search engine position. It's a skill that would have literary purists spinning in their graves.

Among our many target audiences we write copy for tabloids and broadsheets, teenagers, MPs, consumer and professional audiences and now, it seems, we're writing for machines too.