Google review replies written in your own voice
A reply can be perfectly correct and still not be yours. That is true of every one starting with "We sincerely thank you for your feedback": nothing wrong with it, except that an owner writing for themselves never writes like that.
So the question here is not quality but recognition: would someone who knows you read this reply without thinking it is not you?
No commitment · Cancel in one click
Where the tone comes from
| Reply | A usual tone setting | |
|---|---|---|
| Source | Your replies already published on Google | A list of checkboxes |
| Register | Taken from your own text | Chosen from a menu |
| Usual length | Taken from your own text | An average length |
| Sign off | Yours, if you have one | An optional field |
| Things never to say | You list them, they are respected | Rarely provided for |
| Preview before buying | Two replies generated on your real reviews | A demo example |
How the tone is built
The signup funnel asks a few questions: your sector, how you address customers, what you never want to read in a reply. The examples shown in those fields follow your sector, because asking a pharmacist about their services while showing an example about duck breast is a form built for restaurants.
The tone text used by the generator is derived from those answers, never entered twice. If you edit that text by hand it replaces the questionnaire answers: two sources of truth for the same information always end up diverging.
On top of that sits the most useful sample of all: what you have already published on your profile. Your past replies say more about how you write than any description.
What tone does not replace
A language model only knows what it is given. It does not know that the unhappy customer on Tuesday evening is your supplier’s brother in law, or that service was slow because two people were missing.
That is exactly why everything below four stars comes to you by default, with the option to rewrite the reply in one message. Tone automates the form, not the judgement.
What it changes
- Replies your regulars do not find odd.
- Natural variation between two replies rather than a repeated formula.
- Respected prohibitions: what you never want to read does not appear.
- A real preview before any payment, generated with the production prompt.
The limits, stated plainly
- With no reply history the sample is missing, and the tone starts from the questionnaire.
- The model knows neither your customers nor what happened that evening.
- A very distinctive, very spoken or very funny voice gets close but is not perfectly imitated.
- You remain the best writer of your own replies, when you have the time.
How to check it
The preview step generates two replies on your own reviews using the same prompt builder as production. Nothing is simplified for the demo.
You can rewrite any reply by simply answering the message you received: the text you send is published as is, and it adds to what Reply knows about how you write.
Replies never use an em dash, the most recognisable mark of generated text. That is an instruction written into the prompts, not just an intention.
Common questions
Do I need to have replied to reviews already?
No, but it helps a lot. With no history the tone is built from your questionnaire answers, and you adjust it afterwards.
Can I forbid certain words or phrases?
Yes, it is one of the questions asked at signup, and it can be changed later in the business settings.
Does the tone apply to replies in another language?
Yes. The reply is written in the language of the review, with the same register and the same prohibitions.