While the point made by the authors is certainly a valid one, it's a bit sneaky and not very fitting to their overall message that they have the Y-axes on the ngram graphs not 0-indexed. This makes the google results seem more extreme than they in fact are and is a bit of misdirection in itself.
Compare e.g. to the actual ngram viewer which seems to index by 0 per default:
Kind of. The author could fix a lot of their problems with the very prominent dropdown above the graph letting them select the collection— English fiction for example. The long s character can be tricky for OCR, but is not likely relevant to most people's casual use of the tool. I worked on a team that overcame it in a high volume scanning project so they should be able to correct that with software and their existing page images. The plurals criticism is just wrong— you can even do case sensitive searches.
It's not perfect, but it's not useless, and it's not a "lie"— it's just a blunt instrument. Even if the criticism was factually correct, 'proving' that you can't do fine work with blunt instrument is of dubious value.
I think a lot of folks around here are super thirsty to see big tech companies get zinged and when it happens, their fact checking skills suffer.
Compare e.g. to the actual ngram viewer which seems to index by 0 per default:
https://books.google.com/ngrams/graph?content=said&year_star...
https://books.google.com/ngrams/graph?content=said&year_star...