A study of 88,000 Google reviews found AI fingerprints grew far less in customer reviews than in owner replies, while reviews became markedly more glowing.

ASHLAND, MA, UNITED STATES, November 7, 2026 /EINPresswire.com/ -- Since generative AI writing tools arrived in late 2022, local businesses have changed how they write replies to Google reviews much more than customers have changed how they write the reviews themselves, according to new research from Revfer, an Ashland, Massachusetts AI software company.

Revfer analyzed the full public Google review histories of a random sample of 400 businesses in Greater Boston and MetroWest, 100 each of hair salons, restaurants, dentists and auto repair shops, covering about 88,000 reviews written between 2019 and 2026. An earlier report from the same research found that owner replies have roughly doubled since 2019 and, since 2023, increasingly carry markers consistent with AI writing tools.

Customers changed far less:

The em dash, a punctuation mark common in chatbot writing, appeared in 0.2% of reviews in 2019 and 2.9% in 2026. In owner replies, it rose from 0.1% to 4.2%.
Words strongly associated with chatbots, such as “delve” and “testament to,” remain almost absent from reviews.
The typical review has stayed about 140 characters long since 2021.
The exception is frequent reviewers. People who have written 100 or more Google reviews now write reviews about 60% longer than before ChatGPT’s release, while people who have written only one review now write shorter ones.

The larger change is that reviews became more glowing. The share of 5-star reviews rose from 69% in 2019 to 86% in 2026, and the average rating from 4.40 to 4.60. Hair salons rose from 4.58 to 4.79 and restaurants from 4.29 to 4.48. The language changed too: “amazing,” “incredibly” and “genuinely” became more common, while “good” and “nice” declined. The share of reviews with stars but no written text fell from 47% to 24%.

“Owners reached for AI to keep up with replies. Customers mostly still write in their own words,” said Pankaj Aggarwal, founder of Revfer. “That gap matters. A template reply under a heartfelt review is easy to spot, and when most reviews are five stars, every negative one gets read closely. The reply has to sound like the owner.”

Because reviews carry no label of authorship, the study treats these patterns as hints consistent with writing tools, not proof, and notes that phone keyboards and Google’s own prompts may also play a part. All figures are aggregates; no business or reviewer is identified. The full analysis is on the Revfer blog, and the research report with aggregate data is available at doi.org/10.5281/zenodo.23988571.

About Revfer

Revfer is AI software that runs social media and Google review replies for small businesses and startups, without needing prompts. It learns the business once, drafts posts and replies in the owner’s voice, and publishes them on approval. Revfer is based in Ashland, Massachusetts. Learn more about Revfer at revfer.com.

Pankaj Aggarwal
Revfer
+1 407-416-9408
Visit us on social media:
LinkedIn
Instagram
Facebook