House Cleaning Business Flyer: Test ChatGPT Copy, Then Design in Canva

Ask ChatGPT for three flyer versions that differ only in their lead promise (price, trust or time saved), keep the best two, and give each its own QR code and code word. Lay both out in one free Canva template, drop a small batch of each on matched streets, and reprint whichever books more first cleans.

Many house cleaning flyers fail for a plain reason: they say "quality cleaning at affordable prices" and give nobody a reason to ring today. The second reason is that nobody can tell which flyer produced which booking, so the next print run repeats a guess. The order below fixes both: facts first, tracking second, design third, and a drop run as a fair test to settle it.

Follow me on Instagram@sagnikteaches

Give ChatGPT facts and one job per version

"Write me a cleaning flyer" returns the flyer every competitor already has. The model fills gaps with adjectives, so the fix is to remove the gaps: hand it the facts you can prove, the single offer, the length limits, and ask for versions that differ in one way only. The free ChatGPT plan is enough for this. The work is in the brief, not the plan.

Connect on LinkedInSagnik Bhattacharya

Below is a brief for an imaginary two-person house cleaning business. Swap in your own facts and delete any line you cannot back up. Square brackets mark the blanks you fill.

Subscribe on YouTube@sagnikteaches

You are writing copy for a one-page flyer for a house cleaning business that cleans homes, not offices.

Facts you may use (do not add any others):
- Two-person crew; the same two people visit each home
- Typical job: 2 to 3 bedroom home, fortnightly visits or a one-off deep clean
- Offer: $20 off a first clean, claimed by scanning the code or calling [phone]
- Proof I can show: a current insurance certificate; if something is missed we
  return within 48 hours at no charge; rated [rating] on Google ([month and year])
- Streets covered: [three street or neighbourhood names]

Write THREE versions. Each has a headline of 9 words or fewer, a one-line
subheading, exactly three short bullets and a call to action under 10 words.
Version A leads with price. Version B leads with trust (same crew, proof).
Version C leads with time saved for the homeowner.
Keep the offer, the facts and the call to action identical across versions.
Only the lead promise changes. Plain everyday words, no superlatives,
no claim that is not in the facts list.

Here is the kind of answer that comes back. It is illustrative, written to show the typical flaws rather than copied from a real chat.

Three invented details slipped in even with a "no other facts" instruction: "bonded" in A, "two hours" in A, and "eco-friendly products" in C. None was in the brief. This is the normal failure with flyer copy: the model adds the phrases that often appear on cleaning flyers. Delete them, or confirm they are true before they stay. Version B is the closest to the facts you supplied, which is a useful signal about which version needs the least policing.

Two habits improve the output. Ask for six headlines per version and pick one yourself, because the first headline is rarely the best. And paste back your best-performing text message or review quote and ask the model to match its plain wording, since customers describe a clean far more simply than the trade does.

Cut every claim you cannot prove

A flyer is an advertisement, and every claim on it has to survive a customer asking for evidence on the doorstep. Run each claim through a three-column check before anything goes to print.

Advertising rules on claims like these differ from place to place, so check what applies where you trade. Two further points on reviews. Google's own rules say reviews belong to the reviewer, so ask before you quote a customer's words on a flyer. And describe a rating as "on Google" with a date, because a number with no source and no date reads as a made-up one.

Score the versions before you pay for print

Three versions is a fine number to write and a poor number to print. Every extra version splits a limited number of doors, and splitting is what turns a result into noise. Score the three on five questions, 0 to 2 each, and print only the top two. Keep the version you would otherwise have printed as one of the two, because you need a control to beat.

The scoring teaches something beyond the ranking. Version B scores one on the first question because its headline never says "cleaning", so the subheading has to carry that word: change it to "Your home, cleaned by people who know it." A and C tie on seven. A is the kind of price-led flyer rivals already print, so it becomes the control, and B becomes the challenger. C goes back in the drawer for round two.

Set up tracking before any design work

A flyer with no tracking is a donation. Build three ways to tell the versions apart, because people respond in different ways: some scan, some ring, some mention it on a first visit.

1. One QR code per version. Both point to the same booking page, tagged differently. Google's analytics help says a campaign link should always carry utm_source, utm_medium and utm_campaign, and that utm_content separates similar links inside one campaign, so the pair of addresses is simple.
2. One code word per version. Version A says "mention TIDY", version B says "mention CREW". The word also claims the $20, so people have a reason to use it.
3. One question on every new booking: "How did you hear about us?" Write the answer down in the customer's own words.

https://yourdomain.example/book?utm_source=flyer&utm_medium=print&utm_campaign=streets-test&utm_content=a
https://yourdomain.example/book?utm_source=flyer&utm_medium=print&utm_campaign=streets-test&utm_content=b

Canva's QR code generator offers two kinds. A static code is fixed when you create it, so if the link breaks you must print a new flyer. A dynamic code can be re-pointed later without reprinting. Canva says dynamic codes are editable only by the person who made them, and that scan counts and device data (its QR Code Insights) come with Canva Business or Enterprise. On a free or Pro account, use static codes with the tagged addresses above and read the results from your own booking form or web analytics. Free, Pro and Education accounts also get a Canva logo on the code by default, which is harmless but worth knowing before you plan the layout.

If your booking tool cannot record the tags, copy the booking page twice under two addresses, for example /book-a and /book-b, and count submissions on each. It is a clumsy fix, and it works.

A realistic slip to avoid: putting one QR code on both versions "to keep things simple". Nine bookings arrive and no one knows which flyer earned them, so the next print run goes to whichever headline the owner liked more. Tracking set up after printing is tracking not set up.

Keep one sheet for the whole test. A row per enquiry is enough.

Pick a size, then lay out the flyer in Canva

Canva's size guide lists three common flyer sizes, and one of them is the obvious choice for a door drop.

If you work in metric paper sizes, A5 and A4 sit close to the half sheet and standard sizes. Either way, a single offer needs very little room, and a small flyer forces the discipline of one headline and one action. Choose the half sheet unless you have a specific reason not to.

Open Canva, search for "Flyers", pick your size, and start from a template with plenty of empty space. Then arrange the content in this order from the top:

1. The headline, as the largest text on the page, in two lines at most.
2. The subheading, one line, containing the word "cleaning".
3. Your three bullets, with the proof line first.
4. The offer in a plain shape (a circle or a band) so the eye finds it second.
5. The QR code, the phone number and the code word in one block at the bottom, with a clear border of plain background around the code.
6. Your business name and the streets you cover, small, in the footer.

Two settings matter in Canva, and both sit under File, then Rulers and guides. Switch on Show print bleed: Canva adds 0.125 inches (3.175 mm) of bleed on every side for products that get trimmed, so stretch your background colour out to that edge or you may print thin white lines. Then switch on Show margins and keep every word inside them. To add the code, open Elements, search "QR code", and choose "Create static" or "Create dynamic".

Use two fonts at most, one plain colour for text, and one accent colour for the offer. If you add a premium stock photo or graphic, a free account will see "Pay and Download" at the end, so pick free elements if you want the free route. For wider flyer layout principles, the flyer and printed-marketing design tutorial goes further.

What Canva's AI allowance does and does not cover

A cleaning flyer needs almost no AI inside Canva. The template, your text, your own photos, the QR code and the print export are ordinary features. The allowance matters only if you ask Canva to generate an image or a whole design.

Canva's help pages sort its AI tools into tiers. Standard tools are the light ones, Premium tools include "Create an image with Canva AI" and "Generate a design with Canva AI", and Ultra tools (conversational design and video clips) use the most. Canva's pricing table, confusingly, files image creation under Standard instead, so on a paid plan check the usage tracker after your first generated image rather than guessing which pot it came from. The plans differ like this:

Those prices and multiples are from the pricing page as of November 2026, and Canva does not publish a fixed number of uses because heavier tasks eat more of the allowance. To see where you stand, open your profile, then Settings, then Billing, and look under AI usage. For the full comparison across plans, see what the monthly AI allowance covers on each plan.

The practical advice: do the copy in ChatGPT, do the layout without AI, and keep real photos for anything that shows your work. A free account is enough to finish this flyer.

Export a print file and scan a proof before ordering

Download the file as a print PDF. In Canva: Share, then Download, set the file type to PDF, set Presets to Print, open Advanced Settings and turn on Crop marks and bleed (it is off by default). Canva describes the print PDF as 300 dpi with bleed, crop marks and a colour profile choice. The CMYK profile is a paid-plan option, so on a free account you export in RGB. Ask your printer whether they convert RGB files and whether they will send a proof, because bright blues and greens can look duller after conversion. If you print outside Canva's own print service, check your printer's bleed requirement, since Canva itself says other printers may trim differently.

Print one copy on your own printer, or order a single proof, and run this check before the batch:

• Scan each QR code from arm's length with two different phones, one of them an older model.
• Ring the printed number and text the code word to yourself.
• Read the whole flyer aloud, slowly, for spelling and for any claim you cannot prove.
• Hold it at arm's length for three seconds, look away, then say what the flyer wants you to do.
• Check nothing important sits within the margin or near the trim.

A code that scans on a bright screen can fail on paper if it sits over a photo, a gradient or a dark colour. Black on white is dull and reliable.

Run the drop like a test: matched doors, same week

The point of the drop is to make the two versions face the same conditions, so any gap in response comes from the copy and not from the street. There are two ways to do that.

• Alternate doors on the same street. Version A goes to odd numbers, version B to even. This gives the cleanest comparison, though neighbours can compare flyers, which matters little for a $20 offer.
• Matched streets. Choose pairs of streets with similar home types and similar distance from where your crews start, and give one of each pair to each version. In round two, swap the versions, which cancels any street effect.

Deliver both versions on the same days and in the same week, in daylight, and avoid public holidays. Write down how many doors each version reached, because an uneven count will mislead you. Keep the code words valid for at least six weeks, and count late responders.

Respect the rules of the places you deliver to. Rules on letterboxes and mailboxes differ, and a "no flyers" or "no junk mail" notice means skip that door. If you pay someone to deliver, give them a count to hit per version and ask for a photo of the last door on each street.

Keep a drop log so the numbers behind your results are not memory:

Reading the results: 800 flyers, two versions

Here is a full test, with illustrative figures. A two-person cleaning business prints 400 half-sheet flyers of version A (price-led, the control) and 400 of version B (trust-led, the challenger). Both carry the same $20 offer. The print quote is $0.11 a flyer, so $88 for all 800, and delivery takes six hours at a labour cost of $20 an hour, so $120. Total outlay is $208, or $104 per version. After fourteen days, the sheet shows this:

Cost per booking comes next. Version A: $104 divided by 3 bookings is about $34.67, plus the $20 offer gives roughly $54.67 to win each new client. Version B: $104 divided by 6 is about $17.33, plus $20 gives roughly $37.33. Now the break-even. Say you keep $35 of every $140 clean after cleaners' pay, supplies and travel. Version A needs 1.6 cleans from a new client to repay its $54.67, and version B needs 1.1. Both pass quickly if clients stay for several visits, and your own books tell you whether they do.

Now the question the figures cannot settle by themselves: was B really better, or was that luck? Use a rough counting rule. Take the gap between the two counts and compare it with twice the square root of their total. For bookings, the gap is 3, and twice the square root of 9 is 6. The gap is smaller, so call it a lead, not a win. For scans, the gap is 8 against about 11, which also falls short. B is ahead on every line, so it becomes the control for round two, but you have not proved it. Had the bookings been 4 against 14, the gap of 10 would beat the rough threshold of about 8.5, and you could act with more confidence.

That rule is a back-of-envelope check from basic counting, not a full statistical test, and it assumes both versions reached about the same number of doors. Its value is that it stops you printing 5,000 of a flyer on the strength of a three-booking difference.

Round two: keep the winner, change one thing

The next test keeps B as the control and swaps in one challenger that changes a single element. Good candidates, in this order, are the lead promise (try version C), the offer (a free add-on instead of $20 off), the proof line, and the call to action. Never change two at once, or you will not know which one moved the result. Swap the streets as well, so each version meets the other's neighbours.

Each winning headline has uses beyond the flyer: the same words work in a social post, a van sign and the first line of a quote follow-up. For a wider plan around leaflets, local search and referrals, see how to get house cleaning clients with a local plan. If you want to see the same test run for a different trade, the gutter cleaning leaflet tutorial does it for a one-visit service, and the method carries over.

Before you discount your price, check that the first clean still earns something. Quoting by home size gives you the base figure. And once new clients arrive, the cheapest growth comes from keeping them, so read rebooking regular clients before the next print run.

Stop testing flyers when two rounds in a row produce differences smaller than the rough threshold above. At that point the headline is no longer the problem, and your time earns more on the first visit, the follow-up message and the reason a client rebooks.

Questions before you print the first batch

How many flyers do I need to print for a fair test?

No number guarantees a clear answer, because the counts from a drop are small and chance swings small counts a lot. Start with 300 to 500 per version, treat the first result as a direction rather than a verdict, and confirm it with a second drop on swapped streets before you print thousands.

Do I need Canva Pro to make a cleaning flyer?

No. Canva says its flyer maker is free to use, and templates, text, photos, QR codes and the print PDF export all work on a free account. Pro adds the CMYK colour option, five brand kits and a bigger AI allowance, which are nice to have rather than needed.

Should the QR code open my website or a booking form?

Send it straight to the shortest quote or booking form you have, not the home page. Every extra tap loses people who scanned on a doorstep. If your form cannot record where a visitor came from, make two copies of the page with different addresses, one per flyer version.

Can I use AI-generated pictures on the flyer?

Avoid them for anything that implies your own work, such as a picture of your cleaners or a spotless kitchen you never cleaned. Customers can feel misled when the result does not match the picture. Use your own before-and-after photos, or a plain layout with no picture at all.

Is it all right to put flyers in letterboxes?

Rules on what may go through a letterbox or into a mailbox differ from place to place, so check yours before the first drop. Always honour 'no flyers' and 'no junk mail' notices, and consider handing flyers to people directly or leaving them with a shop that agrees.

Further reads

• How to Get Window Cleaning Customers: Leaflets, Local Search and AI — Leaflets, local search and AI for a neighbouring trade.
• Cleaning Business Marketing Ideas: A Month of Posts With ChatGPT — A month of posts to back up the flyer drop.
• Price House Cleaning Jobs: Build a Calculator With ChatGPT — Price the first clean before you discount it.
• Carpet Cleaning Facebook Ads: Write and Test Offers With AI — The same offer-testing habit, applied to paid ads.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Canva flyer maker, flyer size guide, QR code generator, pricing page and help articles on AI usage, print bleed and download file types; Google Analytics help on campaign URL parameters; Google review guidance.