In what may shortly become an ongoing crusade to save aspiring writers from the depredations of automated grammar- and style-checking software, last night I had a look at AutoCrit. They have a 7-day free trial, so don’t believe anything I say; check it out for yourself.

I uploaded the opening chapter from Book II of my impending YA fantasy series. AutoCrit’s web-based user interface is quite nice, I’ll give them that. The results of its critical operations are mixed.

Their customer support desk sends an automated follow-up email when you sign up for the free trial. I took the liberty of replying to that email, just on the off-chance that they might find some feedback from a dissatisfied potential customer useful — and this morning (two days later) I have a reply from the site’s owner/developer, thanking me for my comments and asking whether I might like to chat about their plans for AutoCrit. This is good news indeed, not because it feeds my ego (though of course it does) but because it indicates that AutoCrit is still in active development and may improve in the future.

In case you don’t want to wade through the details, let’s put the bottom line up at the top: A software-based system of this sort may help you improve your writing by letting you see a couple of specific things you will want to tweak. In my case, AutoCrit helped me discover that I use the words “probably,” “maybe,” and “certainly” too often. You might find it useful in a different way. But the challenges facing developers of this sort of software seem nearly insurmountable. Much of what AutoCrit tries to help with, it doesn’t help with.

Below is what I told them in my first email, expanded for clarity. (I have since sent a longer, more detailed critique.) First up, the check for long words:


I can understand it flagging words that are from my fantasy universe (such as “leafstone”) as having three or more syllables, even when they don’t. But it also thinks “seethed,” “squatted,” “desire,” and “horseless” have three or more syllables. Better dictionary access might solve this problem.

On a more serious note, it thinks “had” is a marker for passive voice. This is simply wrong. “Had” is a marker for perfect tense, and is often used in active voice constructions. (A sample sentence, from my uploaded work: “They had succeeded.”) The primary markers for passive voice are “be,” “been,” “was,” and “were.” (And even those are more likely to be used as a copula with a predicate adjective, or as past continuous tense markers, than in passive voice.) Granted, passive voice is often a sign of weak writing; but it’s hard to see how a computer could reliably spot it.

The notion that phrases such as “as if” and “not even” are cliches is simply bizarre.

The search for potential homonym misuse produces so many results [displayed on-screen as red blocks on the words] as to be useless. Nobody is going to go through a long passage trying visually to ignore all of the red highlights on “an,” “two,” “but,” “one,” and “be” in search of a stray usage of “discreet” where “discrete” was intended. (After which they’d have to look it up in any case, since the homonym finder doesn’t offer definitions of the alternative words.)

Arguably, the phrase frequency counter should not flag two-word phrases in which the first word is an article or the second word is “and.” That information is not useful. I had to sift through quite a long list of three-word phrases (such as “the knitting needle”) to find one that I actually think might be worth editing (“gasping for breath”) so it isn’t used twice in the same chapter.

Why the show vs. tell counter should flag the pronoun “it” as indicating telling is rather mystifying.

Now about pacing. The explanation for this feature says, “The paragraphs highlighted in AutoCrit indicate the slow paced paragraphs in your story.” Below is the first paragraph that AutoCrit thinks is slow-paced. It happens to be the very first paragraph in the novel, and I’m not unaware of the desirability of starting off with a bang:

“Everyone else in the railroad car — more than twenty people, of assorted sizes, shapes, and ages — was shouting and screaming in panic. They were pressed against the windows and gabbling about a forest fire, an inferno on both sides of the track, a roaring furnace into the midst of which the train had, suddenly and without warning, plunged. Some of them were choking and coughing, seemingly on smoke. But Kyura couldn’t smell any smoke, and when she looked out the nearest window the countryside lay peaceful, trees and fields green in the summer dawn.”

AutoCrit thinks that’s slow-paced. What can I say?

One might be tempted to assume that AutoCrit is simply looking for long paragraphs and flagging them as slow-paced, but this seems not to be the case. Some long paragraphs are not flagged, yet the following paragraph consists of a single sentence, and it was flagged as slow-paced:

“What she was going to do when she got there, other than get quickly and messily dead, she had no idea at all.”

In sum, I can think of no possible use for the pacing test, whatever it is. The same could be said for the test of uncommon words in fiction, which cavalierly flags such words as blindness, escaped, reward, padded, cheat, and ripped.


I didn’t pull any punches in this email, but I didn’t have to look hard for problems, either. My advice to writers is simple: If you want to learn to write better, learn to write better. Buy a couple of books on prose style, and read every page carefully. Someday we may have software that can be truly helpful, but we’re not quite there yet.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s