Why not just use my five-star ones?

Because a five-star review and a persuasive review are different things. Most five-star reviews say the same warm, general thing, and a stranger reading three of those has had one question answered three times. The ones that work name a worry and answer it: what it cost, how long it took, what they thought before they bought. Your AI scores every review you have against eight things that tend to shift a reader, and hands you ten with a line on why each scored the way it did. You pick three from the ten, and you approve them on a preview before anyone else sees them.

Can my AI just read my Google reviews page?

Only a handful of them, so it is a poor second best. Reading a page brings back roughly the first four thousand characters of it, which on a reviews page is the newest few and nothing else. Pasting them into the chat is the way that works, in batches if there are a lot of them, because the whole pool is what makes the choice worth making. Uploading a file will not do it: your AI can see that a file is there but cannot read what is inside it. However you hand them over, you see the shortlist and approve the final three before anything is published.

Do I need to ask people before I quote them?

Yes, and ask in writing. Putting somebody’s words on your site alongside their name, their trade and their town is publishing information about a real person, and that needs their say-so. It is also a short, pleasant message to send - most people are flattered. You will be offered a draft of it. Nothing goes on the site until they have agreed and you have approved the page.

Make it better4 min read

Work out which reviews to use

You have twenty kind things customers have said and room for three. Most people pick the three that feel nicest to read, which is the one method that reliably picks the wrong ones. The nicest review is usually the vaguest: lovely to deal with, so pleased, would recommend. A stranger deciding whether to spend money with you learns nothing from it.

The prompt

“Here is everything my customers have said about me - [paste them all in, in batches if there are a lot]. I sell [what you sell]. Score them all, give me your top ten with a line on why each one scored, then check every word still matches what they wrote. I will pick three.”

Signed in, Get my prompt fills in your site for you. Otherwise press Copy and fill in the [bracketed] parts. First time? Connect your AI to bernard once: Use your own AI (opens in a new tab).

What actually shifts a reader

What changes a stranger’s mind is not warmth. Eight things tend to do it, and your AI scores everything you hand over against all eight, out of ten, with a line saying which ones each review earned its marks on.

The ones that surprise people:

Doubt first. “I wasn’t sure at first, but...” beats a review that is glowing from the opening word. It reads as somebody telling the truth rather than doing you a favour.
A comparison. A review that mentions what they tried before you, or what they nearly bought instead, does more work than three that only praise you.
Talking to the reader. “You will not get this anywhere else” lands harder than “I could not get this anywhere else.”
One thing, in depth. A review that goes deep on a single benefit beats one that lists nine features. A long list drags down how good the best item on it seems.
Certainty. Always, never, perfect, completely. Hedging words quietly undo a compliment.

How much emotion helps depends on what you sell. If people buy from you for pleasure, strong feeling works. If they buy you to solve a problem, facts with some feeling beats gushing, and beats flat too.

Bring everything, not your favourites

Hand over the lot: emails, Instagram replies, the Google listing, the ones you think are dull. A review that reads flat to you often scores well, because you already know what you are good at and a stranger does not.

Paste them straight into the chat, in batches if there are a lot of them. Attaching a spreadsheet will not work: your AI can see that a file is there but cannot read what is inside it. Pointing at your reviews page is a poor substitute too, because reading a web page brings back only the first part of it, so you get the newest few and none of the rest.

Ten, then three

You get a shortlist of ten, best first, one line each on why. Then, before anybody chooses, the ten are read back against what you actually supplied, word for word. That step is there for a real reason: an AI handed a pile of reviews will smooth a phrase or quietly merge two of them to please you, and you would never spot it. If anything has drifted, you are shown both versions.

Then you pick three, and the rule is that the three carry three different benefits. Not the three warmest. One about the price, one about the wait, one about the result, and between them a stranger has had all three of their questions answered. If two of your picks say the same thing, you will be told which and offered the best alternative.

Before they go up

Ask each of the three, in writing, whether you may print their words, their name, their trade and their town. Say where it will appear. Most people say yes and are pleased to be asked, and a name with a town is what makes a quote read as a person rather than as marketing.

One thing you will never see written on your page: that a review has been verified, confirmed or checked by anybody. Nobody here has spoken to your customer. What was checked is that the words on the page are the words they wrote, which is a different promise and an honest one.

Once you have your three, Add reviews and testimonials covers what happens next: they go on the page exactly as written, laid out to match the rest of your site, landing at the foot of the page unless you ask for them somewhere better, and you approve it before it is live. The seven that did not make the cut are not wasted either. They belong on a case studies or projects page, where somebody already interested will read all of them.

Questions people ask

Why not just use my five-star ones?: Because a five-star review and a persuasive review are different things. Most five-star reviews say the same warm, general thing, and a stranger reading three of those has had one question answered three times. The ones that work name a worry and answer it: what it cost, how long it took, what they thought before they bought. Your AI scores every review you have against eight things that tend to shift a reader, and hands you ten with a line on why each scored the way it did. You pick three from the ten, and you approve them on a preview before anyone else sees them.
Can my AI just read my Google reviews page?: Only a handful of them, so it is a poor second best. Reading a page brings back roughly the first four thousand characters of it, which on a reviews page is the newest few and nothing else. Pasting them into the chat is the way that works, in batches if there are a lot of them, because the whole pool is what makes the choice worth making. Uploading a file will not do it: your AI can see that a file is there but cannot read what is inside it. However you hand them over, you see the shortlist and approve the final three before anything is published.
Do I need to ask people before I quote them?: Yes, and ask in writing. Putting somebody’s words on your site alongside their name, their trade and their town is publishing information about a real person, and that needs their say-so. It is also a short, pleasant message to send - most people are flattered. You will be offered a draft of it. Nothing goes on the site until they have agreed and you have approved the page.