Is "I Write Like" a Scam?

slfisher 0 451 Views

I write like Dan Brown.

At least, that's according to the I Write Like site , which was sweeping the Intertubes yesterday (more than 100,000 in a single day) as people tried to find out which Famous Writer their deathless prose most resembled. (Brown is the critically slammed author of The DaVinci Code.)

The site was put up on July 9 by Coding Robots, a Silicon Valley-based developer, according to the company's blog .

"Currently it analyzes vocabulary (use of words), number of words, commas, and semicolons in sentences, number of sentences with quotation marks and dashes (direct speech)," wrote Dmitry, who claimed authorship.

The site has received some criticisms, with some people wondering whether it's a scam or simply a random result. People who entered text of Famous Writers into the site often found the text attributed to some other Famous Writer -- including, most notably, Famous Writer Margaret Atwood, who found her work most resembled that of Stephen King. "According to the I Write Like analysis, I write like... Ta da! Stephen King! http://bit.ly/bBCMB3 Who knew?" she tweeted .

"“The Raven” is written in the style of William Shakespeare. Shakespeare’s Sonnet 18 is written in the style of Charles Dickens. The first paragraph of Great Expectations is written in the style of P.G. Wodehouse. It seems authors never write in their own styles," one reader of the blog.

In addition, the database of authors -- which thus far is secret -- seems to be slanted toward men, more than one commenter complained. "Okay, so this is fun, but why are all the writers men? There are zillions of equally awesome women writers out there. Most of the writers I matched on have written things I’ve enjoyed, but, seriously, where are the women?" Another systematically tried many female authors and writers of color, noting that they all apparently write like white men.

On the other hand, one person ran a series of Xs through it, only to have the program tell them the author, too, wrote like Dan Brown.

Maybe the program is smarter than we think.

blogger.jpg 18.31 KB

Dani AI

Generated

Short answer: not a deliberate scam, but a lightweight stylometry toy whose results are noisy and often misleading. The mismatches reported in this thread — ’s Dan Brown result, ’s Shakespeare→Agatha Christie, and ’s Bram Stoker — are exactly what to expect from a small, biased author set and a simple nearest‑match method. Treat outputs as entertainment, not proof of authorship.

Practical reproducibility checklist (use as an experiment log): select several canonical texts with known authors; run each text at multiple sample lengths (50, 200, 500, 2,000 words); record every result; then repeat after (a) removing punctuation, (b) replacing rare words with synonyms, and (c) shuffling sentence order. Genuine stylistic signals should be robust across some of those manipulations; wildly different labels show brittle classification or dataset bias.

Why it mislabels: simple classifiers rely on a few surface features and a limited reference corpus, so genre, subject matter, dialogue density, and training-set composition (overrepresentation of particular authors or genders) strongly skew nearest‑neighbor matches. Short samples have high variance; quoted dialogue or many dashes/commas can dominate the score and mask deeper signals like function‑word frequencies or character n‑grams, which are more stable for serious stylometry.

Practical next steps for anyone wanting rigor: run controlled tests (the checklist above), report confusion patterns, and—if attribution matters—use reproducible methods (open corpora, cross‑validation, features such as function‑word counts and character n‑grams, and a simple classifier). The tool is fun for curiosity and sharing, but the thread’s examples make clear it isn’t a reliable authorship detector. 

# quick pseudo-test plan
for size in [50,200,500,2000]:
    sample = take_first_n_words(known_text, size)
    record(size, iwl_predict(sample))
Member Avatar for Xombi
Xombi 0 Newbie Poster

The program is an idiot. I copied and pasted the famous Shakespeare quote "that we call a rose, by any other name" and it analysed the author to be Agatha Christie.

Member Avatar for slfisher
slfisher 0 Posting Whiz

A lot of people have tried a lot of things and gotten a lot of interesting results. :) I wish I could have described them all. Thanks for passing that on.

Member Avatar for php_noob
php_noob 0 Junior Poster in Training

Interesting tool. I paste my writings on my blog and write like Bram Stoker! Cool :)

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.

Edit Preview