An independent used-car dealer can use AI to draft first replies to web enquiries, write listing descriptions from the vehicle's real record, and prepare test-drive follow-ups, so that customers hear back within minutes instead of hours. Keep a person on every price, finance and trade-in statement, because those are the lines that cause disputes.
Each item below comes with a reason and a way to verify it, grouped by three jobs: enquiry replies, listing descriptions and test-drive follow-ups. Work through it once with your own enquiries open beside you. Where an item says "person", it means a named member of staff, not "whoever is free".
Before you switch anything on: the five ground rules
These are set once and cover everything below. Write them on one page and pin it where the sales team can see it.
- AI drafts, a person sends. For the first month nothing goes to a customer without a human pressing send. Review what you are comfortable automating after you have seen a few hundred drafts.
- One source of truth for each car. A single record holding price, mileage, service history, number of keys, known faults and the date of the last check. The assistant only ever sees this record, never your memory of the car.
- A "never say" list. No discounts, no trade-in values, no monthly payment figures, no "no accidents" unless the history report says so, no promises of delivery dates.
- Data goes to a business account. Customer names, numbers and plates stay out of free consumer tools. Use the AI built into your enquiry system or a business plan with sharing off by default, and ask what the vendor stores.
- A log. A shared sheet where staff note every draft they had to correct and why. After a month it tells you which instruction to change.
If you are not sure which platform to use, how to choose an AI platform for a used car dealership covers the buying questions. The checklist works with any of them.
Group one: replying to enquiries faster
Buyers often send the same enquiry to several dealers, and a slow reply gives the others a head start, so this is where speed counts. The goal is a reply that answers the question asked, confirms the car is still available, and offers a next step.
The checklist
- List every place enquiries arrive (listing sites, your own web form, phone messages, social messages, email). Why: AI only helps where leads land in a place you can see. How to verify: put one test enquiry in each channel and time how long it takes to appear.
- Write the reply rules once. Greeting by first name, answer the question, confirm availability from the stock list, suggest two visit slots, sign off with a named person. Why: drafts then sound like one dealership.
- Give the assistant the car record, not the listing page. The listing may be out of date. Verify by changing a record and checking the next draft uses it.
- Handle the five common questions with fixed wording. Is it still available, can I see it today, has it had any accidents, will you take my car in part-exchange, can I get finance. The last three go to a person.
- Set an out-of-hours holding reply. One honest line: "Thanks, we've got your message and [name] will reply when we open at 9." Why: an enquiry that hears nothing until morning has had all evening to go elsewhere.
- Track first-reply time. Write the enquiry time and the reply time in your sheet. Aim to know your median, not your best day.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowA reply prompt you can copy
You write first replies for an independent used-car dealer.
Use ONLY the vehicle record and the enquiry below.
Tone: friendly, plain, short (under 90 words), no jargon.
Always: use the customer's first name, answer the question asked,
confirm the car is available only if the record says "in stock",
offer two visit times from the diary, sign off with [staff name].
Never: quote discounts, trade-in values, monthly payments, or say
anything about accident history unless the record states it.
If the question needs price, finance or part-exchange detail, write
"PERSON NEEDED:" and one line saying what the customer asked.
VEHICLE RECORD:
[paste]
DIARY SLOTS:
[paste]
ENQUIRY:
[paste]What a good draft and a bad draft look like
Say a customer writes: "Is the silver hatchback still there, and do you do 0% finance?" The record says in stock, advertised price $9,495, 61,000 miles, two keys.
A good draft (illustrative): "Hi [first name], thanks for getting in touch. Yes, the silver hatchback is in stock at the advertised price of $9,495, with 61,000 miles and both keys. PERSON NEEDED: finance question, rate and terms. I can show you the car today at 4:30pm or tomorrow at 10:00am. Which suits you? [staff name]". The staff member removes the PERSON NEEDED line, answers the finance question themselves or promises a call, and sends.
A bad draft: "Yes, we offer 0% finance on this vehicle, and we can knock a bit off the price if you can come today." The assistant invented both. This is why the "never say" list exists and why a human checks the draft. For more on scripting these replies, answering a web lead in minutes goes through automating the first response, and ChatGPT prompts for follow-ups and trade-ins has more prompts to adapt.
Group two: listing descriptions that match the car
A listing description should help a buyer decide whether to visit. AI is useful for turning notes into tidy, consistent copy, and risky when it embellishes.
The checklist
- Collect the facts first. Make, model, trim, year, mileage, fuel, gearbox, colour, owners, keys, service history, inspection or roadworthiness date where one applies, tyres, known marks. Why: the description is only as honest as its inputs.
- Note flaws in plain words. "Small scuff, rear bumper, left corner. Stone chips on the bonnet." Why: a listing that admits flaws prevents arguments at the viewing.
- Ask for a description in a fixed shape. A two-line summary, five bullets, one sentence on condition, one on service history. Why: consistency across your stock.
- Read the output against the car. Walk round the vehicle with the draft in your hand. Cross out anything you cannot see or prove, such as "immaculate", "one careful owner" or "full service history" when the book has gaps.
- Check price and finance lines yourself. A person types the advertised price and any finance example; the assistant never does. Why: a mismatch between the page and the forecourt can breach advertising rules, so confirm what applies to you with a qualified adviser.
- Keep a banned-words list. "Perfect", "flawless", "bargain", "guaranteed" and anything that implies an outcome you cannot guarantee.
A filled-in example
Illustrative notes: "2017 hatchback, 1.2 petrol, manual, silver, 61,000 miles, two owners, two keys, service book stamped to 54,000 miles, new front tyres, small scuff rear bumper, stone chips bonnet, inspection valid to next spring."
Illustrative output after editing: "Two-owner silver hatchback with 61,000 miles and two keys. Petrol, manual, economical for town and motorway. Service book stamped to 54,000 miles; new front tyres fitted. Cosmetic marks: a small scuff on the rear bumper and light stone chips on the bonnet. Inspection valid to next spring. Viewing and test drive by appointment." The draft first said "full service history" and "excellent condition". The first was false (the last 7,000 miles are unstamped) and the second was a judgement the notes did not support. Both were removed.
For a fuller treatment of photo captions and clean-up as well as text, honest listings and photo clean-up for used car dealerships is the companion tutorial.
Group three: test-drive and follow-up messages
The quiet days after a viewing are where interest fades. A short, timely message keeps the car in the buyer's mind, and AI can personalise it from your notes in seconds.
The checklist
- Write a two-line note straight after each test drive. What they liked, their stated worry, what they said they would do next. Why: AI cannot invent this and generic follow-ups read as generic.
- Send the same day. A short thank-you with one fact they cared about: "You asked about boot space, it measures [figure from the record]."
- Follow up on a schedule. A message at two days and at seven days, then stop. Why: more feels like pressure, and none feels like indifference.
- Offer something useful, not a push. A video walk-round, a link to the service history, a hold for 48 hours if your policy allows it.
- Treat objections with a person. "I found it cheaper elsewhere" and "my partner isn't sure" are conversations, not templates.
- Stop when told to stop. Record opt-outs in your system and honour them. Check what consent your marketing messages need with a qualified adviser.
Example: a same-day follow-up and a day-seven nudge
Notes (illustrative): "Liked the quiet engine. Worried about insurance cost as a first-time buyer. Said he'd talk to his dad at the weekend."
Same-day draft: "Hi [first name], thanks for coming in today. You asked about insurance for a first car, so I've put the year, engine size and exact trim from the car's record below, ready for when you ask for quotes. No rush. If it helps, we can hold the car until Monday at 5pm. [staff name]". Seven-day draft: "Hi [first name], just checking whether you got your quotes sorted. The hatchback is still here if you'd like another look, or I'm happy to answer anything your dad wants to ask. [staff name]".
The detail that makes it work is the note. Without "worried about insurance" the assistant has only generic content. For more on handling a customer's old car in the conversation, trade-in enquiries before the appraisal shows what to collect first.
A worked week for a lot of 25 cars
Say a small lot holds 25 cars, receives about 14 web enquiries a week, books about six viewings and sells about three cars. The owner and one salesperson handle all of it. These are illustrative numbers to show where time moves, not results to expect.
| Task | Before (minutes a week) | With AI drafts and checks |
|---|---|---|
| First replies, 14 enquiries | 14 x 6 = 84 | 14 x 2 = 28 (read, fix, send) |
| Listing text, 4 new cars | 4 x 25 = 100 | 4 x 10 = 40 (notes, draft, walk-round check) |
| Follow-ups, 6 viewings x 2 messages | 12 x 5 = 60 | 12 x 2 = 24 |
| Logging corrections | 0 | 15 |
| Total | 244 | 107 |
A saving of about two hours a week is modest, and that is a realistic scale for a lot this size. The more important change is that the 14 enquiries each get a reply within an hour or so, including the ones that arrive in the evening and are answered at opening time with a holding reply. Whether that wins extra sales is something to measure in your own numbers over a quarter.
Where it goes wrong on a forecourt
- Stale stock. The assistant says a car is available after it sold on Friday. Fix: the record has a status field updated the moment a deposit is taken.
- An invented incentive. A draft offers "a free service" the dealership has never offered. Fix: the "never say" list, plus a person reading each draft.
- Wrong plate or customer. A draft mixes up two enquiries. Fix: one enquiry per prompt, never a batch of several customers in one paste.
- A mismatch between ad and showroom. The listing says 61,000 miles and the car now shows 61,150 after a delivery run. Fix: update the record, then regenerate, and check mileage on the day of sale.
- Tone drift. After a month the drafts start sounding like a template. Fix: refresh the instruction with three recent replies you liked.
Four awkward messages and how to handle each
The checklist covers the routine. These four situations come up weekly and are where a drafted reply either saves you or embarrasses you. In each, the assistant drafts, and a named person decides.
The car has just sold
An enquiry arrives for a car that went on Friday. The wrong reply says it is available. A good prompt tells the assistant to check the status field first and, if the car is sold, to suggest up to two similar cars from stock using only the records. Illustrative draft: "Hi [first name], thanks for asking. That hatchback has just sold, sorry. Two similar cars are in stock: a 2016 model at 58,000 miles and a 2018 model at 66,000 miles. Would you like details of either?" The person checks that both suggestions are really similar in price band and then sends.
The customer wants the price down
The draft must not negotiate. The instruction is a single line: "If the customer asks for a discount, write PERSON NEEDED and a one-line summary." The salesperson then replies by phone or message in their own words. The assistant's only contribution is to flag it quickly and prepare a one-line note of the car's days in stock, which the person already knows but may not have to hand.
The customer asks about history you do not have
A buyer asks whether the car was ever a rental or ever written off. If the record has no answer, the draft should say so plainly: "I don't have that detail in front of me, and I don't want to guess. I'll check the history report and come back to you today." Then a person actually does it. The failure to avoid is the assistant filling the gap with a reassuring sentence.
The angry message
A customer complains about a fault after purchase. This is not a drafting task. Have the assistant do one thing only: summarise the message in two lines and list the dates mentioned, so the owner can reply properly. The reply itself, and any decision on repair or refund, belongs to the owner, and wording about legal rights should be checked with a qualified adviser.
A one-page enquiry sheet to give every salesperson
Print this and keep it by the keyboard. It turns the checklist into a ten-second routine for each new enquiry.
- Open the vehicle record. Is the status "in stock"? If not, use the sold-car reply.
- Paste record, diary slots and enquiry into the reply prompt. One customer at a time.
- Read the draft. Is every fact in it on the record? Is there any price, finance, part-exchange or history line? If yes, replace it with your own words or call.
- Check the greeting name and the signature name.
- Send. Put the time in the sheet.
- If you changed more than a line, write one sentence in the corrections log saying why.
New staff pick this up in a day because the sheet is short and the prompt does the heavy lifting. Review the corrections log every Friday for ten minutes. If the same fix appears three times, edit the prompt, and date the change so you can see whether the corrections fall afterwards.
Measure it for four weeks
Keep four numbers in a sheet. Median minutes to first reply. Enquiries that became viewings. Viewings that became sales. Drafts corrected by a person, with the reason. After four weeks compare against the month before. If reply time dropped and the correction count fell, move to the next job, perhaps a small website chatbot; the rules to set before a chatbot launches covers how to do it without losing control. If the correction count is high, rewrite the instruction before you add anything new.
A final word on scope: this checklist keeps AI to wording and speed. It does not value cars, approve finance, or state vehicle history. Those stay with trained staff and your documents, and wording that touches advertising or credit should be checked against the rules that apply to you. If you want help mapping the leads and the tools you already have, that is what my AI implementation consultation covers, through the contact page.
Questions from dealers starting with AI
Should AI quote a price to a customer?
Only the advertised price that is already on the listing, and only as a copy of it. Anything that involves a discount, a trade-in value, finance terms or monthly payments should come from a person, because a wrong figure sent in writing can be treated as a promise.
How fast should a dealer reply to a web enquiry?
As fast as you can reliably sustain. A drafted first reply within minutes beats a perfect one the next morning, but only if someone checks the facts. Pick a target you can keep during busy Saturdays, then measure it for a month.
Can I paste customer details into a chat assistant?
Avoid it on a free consumer account. Use a business plan with training switched off by default, or strip names, phone numbers and registration plates first. Your enquiry system may also have built-in AI that keeps data inside it, which is usually the safer place.
Will AI-written descriptions hurt my listings?
Not if every claim comes from the vehicle's real record. Generic descriptions that sound the same for every car are weaker than short, specific ones. Give the assistant the actual spec, condition notes and history, and delete anything it adds that you cannot see on the car.
Further reads
- Can an AI Voice Agent Book Car Dealership Service Appointments? — Whether an AI voice agent can book service visits.
- AI BDC for Car Dealerships: Does a Small Lot Need One? — Whether a small lot needs a dedicated enquiry desk.
- Responding to Dealership Reviews With AI: Samples and Rules — Sample replies to customer reviews.
- AI Marketing for Car Dealerships: A Four-Week Plan for One Lot — A four-week marketing plan for one lot.
- Mia AI for Car Dealerships: Questions to Ask Before You Sign — Questions to ask a vendor before you sign.
- Car Dealership AI Receptionist: Monthly Cost and Hard Limits — What an AI receptionist costs and cannot do.
- ChatGPT for Car Sales: Follow-Ups, Listings and What Never to Promise — Follow-ups, listings and trade-in replies drafted with ChatGPT, plus the never-promise list, a rules block to paste and a listing audit.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: general dealer practice and the author's guidance; OpenAI's enterprise privacy page and Anthropic's commercial terms on business data and model training (read 5 November 2026); no vendor pricing is quoted. Check your market's advertising and consumer-credit rules before publishing price or finance wording.