Landscaping Business AI: Review Requests Built From Job Photos

Photograph the finished job, have an AI assistant describe only what the photos show, add the two or three facts photos cannot show, then ask it to write a short personal review request that includes your Google review link. Send the same kind of request to every customer, offer nothing in return, and read each message before it goes.

Personal beats generic only if the personal part is true. An assistant can describe a patio from a photo, but it will also guess at materials, sizes and who lives there, and a request that gets those wrong is worse than the standard text from your software. So the method keeps you in charge of the facts and uses AI for the wording.

Follow me on Instagram@sagnikteaches

The worked example is one patio job, followed from the last day on site to the sent message, with the prompts, the output and the corrections. Google's rules decide what that message may say, so start with them.

Connect on LinkedInSagnik Bhattacharya

What Google's rules allow a review request to say

Google's contribution policy sets limits that apply to every message you send, however it was written. The wording I would build into every prompt:

Subscribe on YouTube@sagnikteaches

• No incentives. Offering payment, discounts, free goods or services in exchange for posting a review is not allowed. A 10% off voucher "when you post" is an incentive.
• No gating. The policy bars discouraging negative reviews or selectively soliciting positive ones from customers. Asking only the customers you think are pleased is selective.
• No scripts. Merchants should not request that specific content be included, and the policy specifically rules out asking staff to solicit reviews with specific content or a certain number of reviews. So no "please mention the drainage work" and no "mention Sam by name".
• No pressure. The policy says merchants should not require or pressure people to leave ratings or reviews while on the premises. For a trade business, the sensible reading is no asking for a review on the doorstep at handover.

Two practical points sit alongside the rules. A customer has to be signed in to a Google Account to post, though Google's help page says it can be created with a non-Gmail email address. And if you send the request by text, send it only to customers who have agreed to be contacted that way. Marketing-message rules differ between places, so ask an adviser if you are unsure.

The job we follow: a porcelain patio, a planter and a drain

Illustrative job: a 32 m² porcelain patio outside a family's kitchen doors, a raised timber planter along the back fence and a slim drainage channel beside the house wall. Three people, five working days. The dig uncovered a blocked old land drain under the kitchen end, which took an extra day to clear and re-lay. The customer's first name goes in as [first name], and the crew lead is [your name].

The final visit was a Friday. The photos are taken that afternoon, the caption and job note are written on Monday morning, and the request goes out that afternoon, three days after the crew left. Spacing it keeps the message from arriving while the customer is still sweeping up dust.

Five photos an assistant can read, and what to keep out of them

1. A wide shot from the kitchen doors looking out over the whole patio.
2. The same space from the far corner looking back at the house.
3. A close-up of the paving surface and joints.
4. The drainage channel and the edge where patio meets wall.
5. The planter in full, including the plants.

Shoot in daylight, level, with the tools put away. Take them from the same positions as your before photos so the pair is easy to compare later. Keep people out of the frame, and watch the edges for house numbers, number plates, children's belongings and anything on a table. Phones can store the location of each photo, so check the share setting that leaves it out before you send photos anywhere or publish them.

Step 1: turn the photos into a caption you can trust

The first prompt asks for description only, with the assistant admitting uncertainty. Attach the five photos.

These are five photos of a finished landscaping job. Describe only what you can
see, in plain sentences, as a caption for my own records.
For every detail say "clear" or "guess". Do not state materials, sizes, plant
names or directions unless they are plainly visible; write "unclear" instead.
Do not describe any person.
At the end list anything in the photos I should crop or remove before sharing
(house numbers, number plates, faces, personal items).

Illustrative output: "A paved terrace of large rectangular light-grey tiles in a straight grid with narrow joints (clear), beside a house with glazed double doors (clear). The tiles look like limestone (guess). A slim drainage channel runs along the house wall (clear). A raised bed of timber with grasses and shrubs sits along a fence (clear); the timber looks like cedar (guess). The terrace is about 40 square metres (guess). Remove: a house number visible in photo 2; a child's scooter in photo 5."

What to fix: the assistant labelled its own guesses, which is the point of the prompt, but you still have to correct them. The tiles are porcelain, not limestone. The timber is treated softwood sleepers, not cedar. The terrace is 32 m², not 40. Crop out the house number and the scooter. If the caption were only going to be reused, the "guess" labels would have been worth the extra words, because they show you which lines to check.

Step 2: add the facts a photo cannot show

A short job note is what turns a description into something personal. Three lines are enough, and they come from you or the crew lead, not from the assistant.

What they asked for: a flat dry place to eat outside that the children can also
use for scooters; a planter instead of the old borders.
What changed: we found a blocked old land drain under the kitchen end, cleared
and re-laid a short section, and added a day.
One practical detail: the patio falls gently away from the house so water
runs off, with the channel catching what reaches the wall.

The job note is also the check on step 1. If the caption says one thing and the note says another, the note wins. This takes about three minutes and is the part nobody can automate, because only the people who were there know it.

Step 3: write the request, with the rules inside the prompt

Now give the assistant the corrected caption, the job note and the rules, and ask for a message.

Write a review request text message from a landscaping business to a customer
whose patio we finished last week. Use the corrected caption and job note
below. Tone: friendly, plain, under 90 words, no emojis.
Rules: do not ask for a star rating. Do not suggest what the review should say
or ask the customer to mention any person. Offer no discount, gift or reward.
Say the review is welcome whatever the customer thought. Refer to the job in
one personal sentence using only facts from the caption and note. Include the
placeholder [review link]. If anything was not right, invite them to tell me
directly as well. Address them as [first name] and sign as [your name].

A first draft without those rules tends to look like this, and it breaks three of Google's rules at once. Illustrative:

Hi [first name]! We'd love a 5-star review on Google. Please mention how professional Sam and the team were. As a thank you, we'll take 10% off your next maintenance visit when you post. [review link]

It asks for a rating, asks the customer to mention a named staff member and offers a discount for posting. Illustrative output from the prompt with the rules inside it:

Hi [first name], it's [your name] from [business]. Thanks again for having us do your patio and planter. It was good to leave you with a flat, dry terrace outside the kitchen doors after the drain surprise, and I hope the scooters are getting some mileage. If you have a few minutes, a review on Google would help other people find us, and it's welcome whatever you thought: [review link]. If anything isn't right with the job, tell me directly as well and I'll put it right. Thanks, [your name]

Read it as the customer would. The sentence about the drain and the scooters is the part a template cannot produce. The line "if anything isn't right" sits beside the request rather than replacing it, so it does not steer unhappy customers away from reviewing. Change anything that does not sound like you: if you never write "a few minutes", cut it.

Step 4: find your Google link and send the message

Google's help page for the review link covers the browser route. Open your Business Profile, choose Read reviews, then Get more reviews, select the share icon and copy the link. The QR code can only be made in a computer browser, not on a phone; right-click it to save the image. The page suggests sending the link by email, WhatsApp or a Facebook post, and putting it on receipts and thank-you emails. Its sentence on rules is the same one as above: a review has to reflect a genuine experience and nothing may be offered for it. The help page is short and worth reading once.

Before the first send, open the link on a phone that is not signed in to your business account, and check it takes you to the right profile. Then replace the [review link] placeholder with it and send from your own phone or business number. A message from the person who stood in the garden reads as personal in a way a no-reply address cannot, so send it from a person.

The same method on a smaller job: a hedge and a path

The patio had plenty to photograph and plenty to say. Most jobs do not. Take a customer on a monthly maintenance plan whose hedge was cut back hard, whose gravel path was re-edged and whose borders were weeded. You take two photos, the path in full and the hedge line. The caption prompt returns two sentences and labels one guess: the hedge "looks like laurel (guess)". It is privet, so the caption is corrected before it goes anywhere.

The job note has one line: the hedge was about two metres wide at the top, you took it back by roughly 40 cm so the path is usable again, and the customer had mentioned the path was getting hard to walk on. The prompt from step 3 then produces this, illustrative as before:

Hi [first name], it's [your name] from [business]. The path should be easy to walk again now the hedge is back where it belongs, and the edging is crisp. If you'd like to leave a review on Google it would help other people find us, and it's welcome whatever you thought: [review link]. If there's anything you'd like done differently on the next visit, just say. Thanks, [your name]

A small job earns a shorter message, but the personal sentence still does the work. If a job gives you nothing personal to say, because it was a routine visit and nothing changed, send the standard request from your software and do not force a line. A strained compliment about a lawn is worse than none.

Every customer gets one, not only the happy ones

The method falls apart if it becomes a filter. Write your rule down: every completed job gets one request, two to three days after the last visit, to the person who signed it off. Keep the rule where your crew leads can see it.

The awkward case is a customer with an open complaint. Hold the request while the complaint is open, fix the problem, and send it when the matter is closed, whichever way it ended. Do not remove them from the list because of the complaint. If you ever need to show how you work, a written rule applied to every job is the best evidence you have that you did not select by sentiment.

Also keep staff quotas out of it. A crew lead who is told to collect a certain number of reviews will start choosing customers, and Google's policy lists requesting that staff solicit a certain number of reviews as something merchants should not do.

Automation for the standard request, AI for the personal one

If your job software already sends review requests, use it for the routine ones. Jobber's help article on Reviews says its tool sends automatic requests for a Google review, needs a verified Google Business Profile, and is part of its Marketing Suite, which Jobber's pricing page lists as a $99-a-month add-on (included on Plus). Triggers include a job closing, a visit being completed or an invoice being paid in full. It sends at most one request per client every six months, sends by email only if the trigger falls between 9pm and 9am, and follows up twice, three and five days later.

That is a good baseline for a lawn visit. For the jobs worth a personal note, such as a patio, a full planting scheme or a front garden redesign, the photo method earns its ten minutes. Pick one route per job: do not let the software and your personal message both go to the same customer. My tutorials on automated Google review requests and review follow-ups for trades cover the automatic side in more depth.

Reusing the photos and caption on your profile

The same photos can go on your Google Business Profile once the customer agrees. Use the real photos: Google's photo tips say an image should have no significant alterations or excessive use of filters or AI, and should represent reality, so do not generate an "after" image. The corrected caption becomes the post text, and the profile lets you schedule posts. My tutorial on getting landscaping clients from local search sets out a four-week plan that includes this. Keep the order: customer permission first, then crop, then publish.

What it costs in time and money, and a four-week start

Budget about ten minutes a job: three for the photos, two for the caption, three for the job note and two to read and send. On twelve finished jobs a month that is two hours. If you have a ChatGPT or Claude subscription, the extra cost is nil, since individual plans list at about $20 a month (Claude Pro is $17 a month on annual billing). A free plan may be enough to try the method on a few jobs, but image uploads are capped, so check the current allowance before you rely on it.

1. Week 1. Write the rule (who gets a request and when), save the three prompts in a ChatGPT Project or a Claude Project so nobody retypes them, and run the method on three jobs from last month without sending anything.
2. Week 2. Switch on for live jobs. Crew leads take the five photos at the end of the last visit; you write the caption and note, and send the message.
3. Week 3. Start the tracking sheet below. Check every link and read each message aloud before it goes.
4. Week 4. Look at the log and at any customer reactions. Adjust the prompt wording, then decide whether crew leads can run steps 1 to 3 themselves with you reading the final message.

A tracking sheet that shows whether it works

A review request is easy to send and easy to forget, so log each one. Eight columns are enough. These rows are made up to show the layout, and the numbers are not benchmarks.

After 30 or 40 requests of each style, compare how many of each produced a review within two weeks. Ten requests tell you nothing, and I would not change your approach on fewer than 30. The number that matters more is whether anyone said the message felt pushy. If one customer says it, soften the wording; if none does, leave it.

When the review arrives: thank them without repeating private details

Reply to every review, briefly. Say thank you, refer to the job in general terms, and leave out addresses, prices and anything the customer did not write themselves, such as the drain problem. An assistant can draft the reply from the review text and your job note, and you post it. For reply wording, including how to handle a bad review, see my tutorial on using AI to respond to Google reviews. Keep the reply as short as the review: a four-line review does not need a paragraph back.

Three ways a photo-built request goes wrong

The caption states a wrong detail with confidence. An assistant that writes "limestone" or "cedar" when the material was porcelain or treated softwood has put a false statement into a message with your name on it. The "guess" labels and the job note are the defence, and so is reading every message.

The photos carry more than you meant. A house number or a child's toy in a photo sent to an AI chat is information you did not need to share, and if the output mentions something the customer never told you ("I hope the kids enjoy the new trampoline") you have crossed from personal to intrusive. The scooter line in the patio message works only because scooters were in the customer's own brief. Crop first and give the assistant only the job note's facts about the household.

The link points to the wrong place. A business that has moved or renamed can end up with two profiles, and customers may review the old one. Test the link on a phone that is not signed in to your business account before the first send, and again whenever you change your profile. For the wider picture of how enquiries, jobs and reviews connect, see the whole-business workflow map.

Questions about AI-written review requests

Can I offer a discount for leaving a review?

No. Google's contribution policy lists offering payment, discounts or free goods and services in exchange for posting any review as prohibited, and its review-link help page repeats it. That includes a thank-you voucher promised in the request. Thank customers afterwards in a reply if you like, but never tie anything to posting.

Can I send requests only to customers who seem happy?

No. Google's policy prohibits selectively soliciting positive reviews or discouraging negative ones. Decide a rule before the job, such as one request to every customer two to three days after the last visit, and apply it every time. If a complaint is open, resolve it first and then send the request, whichever way it ended.

Do customers need a Gmail account to leave a Google review?

They need to be signed in to a Google Account, but Google's help page says they can create one with a non-Gmail email address. That is worth a line in your message for customers who say they cannot find where to post, because the account step is where most of them stop.

Is it safe to upload customer job photos to an AI chat?

Only photos that do not identify anyone: no faces, house numbers, number plates or children's belongings. Business plans from ChatGPT and Claude do not train on your content by default; consumer plans have a training switch in privacy settings that I would turn off. Ask the customer before you publish any photo of their property.

Further reads

• Should You Let AI Auto-Post Replies to Google Reviews? — Whether AI should post replies to reviews for you.
• HVAC Google Business Profile: Service Areas, Q&A and Review Replies With AI — Profile posts and review replies for another trade.
• AI Marketing Video Generator: Before-and-After Clips for Landscapers — Turn the same before-and-after photos into short clips.
• AI in Landscape Maintenance: Contract Renewals, Visit Notes, Upsells — Ask for reviews from maintenance customers at renewal time.
• Best AI for a Landscaping Business: Eight Tools for Quotes, Calls and Crews — Eight tools for a landscaping firm, sorted by job, with list prices, peak-month call-cost sums and separate stacks for one crew and five crews.
• Landscape Business AI Assistant: Build One for Quotes and Client Replies — Set up a ChatGPT or Claude Project that drafts landscaping quotes and replies from your price list, with ten test enquiries and handover rules.
• Appliance Repair Marketing: Local Pages and Review Requests With AI — A job-log test for which brand-and-fault pages to publish, a fact sheet that stops AI inventing, and review requests that follow Google's rules.
• Bathroom Fitter Marketing: Before-and-After Posts From Job Photos With AI — A bathroom fitter's routine for matched photos, customer permission, AI-written captions, Google Business Profile updates and a one-month posting calendar.
• Eight Lawn Care Marketing Ideas for Spring, Written With AI — Eight spring lawn care campaigns with prompts, sample messages, a worked pre-booking plan, a February calendar and the email and text consent checks.
• Lawn Care Letter to Customers: Welcome and Delay Notes With AI — Welcome, delay and season-end letters for a lawn care business, drafted with AI from a facts sheet, with tone settings and checks before sending.
• Paving Marketing: Turn Job Photos Into Local Content With AI — A six-shot routine, captioning prompts and permission wording that turn each paving job into posts, a project page and a profile update.
• Pressure Washing Marketing Ideas: Neighbour Offers and Posts With AI — Five neighbour offers with wording and sums, a 13-minute photo-to-post routine, Google Business Profile rules, and how to get permission for job photos.
• Roofing Social Media: Turn Finished Roofs Into a Month of Posts With AI — A consent step, a one-line job note and a four-week calendar that turn finished-roof photos into sixteen honest posts.
• Winter Work for Landscapers: Plan Off-Season Services With AI — A winter service menu for landscapers (snow, lights, pruning, planning) and a way to test demand with past customers before buying equipment or hiring.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Google Maps user-generated content policy (prohibited and restricted content); Google Business Profile help (review link and QR code, photo tips); Google Account help (creating an account with a non-Gmail address); Jobber help centre (Reviews) and pricing page; read on 1 November 2026.