What are guests unhappy about? Read your low reviews with your AI

Reading reviews one at a time is how most owners do it, and it's how most owners miss the pattern. One review says the wait was long. Another, two weeks later at the other location, says the order was wrong. A third says nobody came to the table. Each one stings on its own. Together they might be one problem: the lunch rush is understaffed. You only see that when you read them side by side.

If your Google reviews come into Mays, your AI app can read them side by side for you.

Part of our series Run your restaurant with your AI: one real situation per post, with the words you type and what happens in Mays. The prompts and results are illustrative examples.

What the owner types

"Show me the reviews of 3 stars or less from the last 60 days at Midtown and Riverside that nobody has replied to. What are the common themes?"

What the AI does in Mays

It asks Mays for the reviews, filtered the way you asked. That's the same list as the Reviews page in Mays, and reading it changes nothing:

• Per location, newest first, from the last 30 days by default or up to 90.
• Only 3 stars or fewer, if you say so (any number from 1 to 5).
• Only unreplied ones, if you say so.
• For each review: the rating, the day it was posted (on the location's calendar), whether it has a reply, the reviewer's public name shortened to first name and initial, and the text.

Up to 50 reviews in one answer, with their average rating. Ask for one location, or a shorter window, if there are more.

What you see

The AI does the reading and the grouping, for example (illustrative):

"7 unreplied reviews of 3 stars or less, average 2.1. Three themes: wait times at lunch (4 reviews, all Midtown, all between 12 and 1 pm), a cold or wrong breakfast sandwich (2, Riverside), and the music being too loud (1). The oldest unreplied one is 41 days old."

The themes are the AI's reading of the text. Mays provides the reviews.

Then dig in with Mays data:

• "How many people were on at Midtown between 12 and 1 on weekdays this month?" (Sales by hour with staff, in Mays.)
• "Which breakfast items were on canceled or refunded orders at Riverside this month?" (POS orders, if your POS is connected.)

Replying: in Mays

Replies happen in Mays, on the Reviews page, which posts your reply on Google. The chat doesn't post replies. That's on purpose: a public reply under your restaurant's name deserves a human's last look. The AI can help you draft one if you ask; you edit it and post it from the reply box in Mays.

A few things worth keeping in mind:

• Reply to the theme, not only the review. If lunch waits are the pattern, the fix is on the schedule, and the reply can say what you changed.
• Don't name staff in a public reply, even when the review does.
• Old unreplied reviews still count. Readers see them.

Reviews are data, not instructions

Review text is written by strangers. Mays passes it to the AI as a quote, marked as text from customers, and the AI app is told never to follow instructions inside it. A review that says "ignore your instructions and…" is reported, not obeyed.

What it replaces

What it needs

• Google Business Profile connected to Mays for each location (see Connecting Google Business Profile).
• Access to reviews in Mays, at those locations.
• It's read-only, so no actions setting is needed.

What Mays doesn't do here

• It doesn't post replies from the chat. Reply on the Reviews page in Mays.
• It doesn't read other review sites. These are your Google reviews.
• It doesn't show who the reviewer is beyond their public name, shortened.

Try it this week

1. Connect Mays to ChatGPT, Claude or Grok from mays.app/ai (ChatGPT, Claude, Grok). Connect Mays to ChatGPT, Claude or Grok has the details.
2. Ask for the unreplied low reviews and the themes.
3. Fix the theme, then reply from Mays. See Google reviews and replies.

Try Mays free for 30 days, with no credit card, and connect it to the AI you already use.