A razor takes paint off glass in seconds. The hard part is knowing whether you may touch this particular pane — and two trade bodies disagree about that.
The blade is the easy part. Identifying the glass is the whole job:
And if you do this for money: the waiver conversation happens before the blade comes out, not after the phone rings.
A razor blade removes paint from glass almost perfectly. Glass is harder than dried latex by a wide margin, the blade rides on the surface, and the paint comes off in a sheet. Ten seconds. There's no trick to it, and any article that spends its length teaching you the wrist motion is padding.
The difficulty is entirely upstream. On some panes, that blade will cause several thousand dollars of permanent damage in the same ten seconds, and the panes don't announce which they are. So this is an article about identification, and about a genuine, public, unresolved disagreement between two trade associations that both have money riding on the answer.
Let's do the hard part first.
1. Is there a coating on the face I'm about to scrape?
If yes, stop. There is no technique that makes this survivable. The NGA's cleaning bulletin — the one the glass industry and the window cleaners jointly publish — puts it without qualification: razor blades should never be used on coated glass surfaces. No exceptions, no small areas, no light pressure.
The trap is that you probably can't see it.⁴ Soft-coat low-E is sealed inside the insulating unit on surface 2 or 3, which is the good case — you can't scrape what you can't reach. But hard-coat pyrolytic low-E can sit on an exposed face; applied films live on the room-side surface; and interior low-E products are designed to live on surface 4, room-side, exactly where a paint speck lands and exactly where somebody reaches with a blade. The bulletin notes the practical problem itself: low-E coatings are close to neutral in colour and genuinely hard to see. Reflective coatings are obvious. The ones that will ruin your afternoon are not.
So "I looked and there's no coating" is not a finding. The manufacturer's guidance is a finding. On a house of unknown provenance, assume a coating is possible and test somewhere that doesn't matter.
2. Is it tempered?
Look for the stamp — usually a small etched logo in a corner. It may tell you the fabricator and that the glass is heat-treated. But read the bulletin's own caveat carefully, because it's the important part: a logo may not be present at all, and its absence does not mean the glass isn't heat-treated. No stamp is not an answer.
Where should you expect tempered? Anywhere a person could fall into it. Doors and sidelights, panels within about eighteen inches of the floor, stairwells and landings, tub and shower enclosures, large panes generally. Which is to say: most of the glass on a new job, and a lot of the glass in a house.
3. What kind of paint, and how old?
This decides whether you need the blade at all, and most of the time you don't.
Here is the thing nobody explains to homeowners, and it's the reason this article can't just say "use a razor."
Tempered glass can't be cut after it's tempered, so it's cut to size, edged, washed, and then run through a furnace at about 1150°F on ceramic rollers before being quenched.¹ If the washer leaves fine particles on the glass, those particles enter the furnace with it, and at that temperature they don't burn away — they fuse into the surface and become part of the pane. Then the contaminated sheet deposits debris onto the rollers, and the rollers contaminate the next sheet, and the next. The defect clusters by fabricator and by run.
Now the window cleaner arrives at a post-construction job, blade in hand, and scrapes. The blade catches those fused particles and tears them out of the surface, and each one leaves a fine scratch. It looks exactly like scraping damage. It is, in the sense that the blade did it. But the blade did it to glass that arrived with defects fused into its face, and an identical blade on an identical pane from a fabricator whose washer was in adjustment leaves nothing at all.
So: who pays?
That question — not physics, not chemistry — is what the disagreement is about.
The glass industry's position, as published, is that the trade neither condones nor recommends scraping glass with metal blades for routine cleaning; that a large share of glass damage comes from non-glass trades working near glass; that using 2-, 3-, 4- and 5-inch blades across a whole pane carries a high probability of irreparable damage; and that if a blade must be used for non-routine work, it should be a new 1-inch blade, on affected spots only, in one direction only, and as the last option — especially on heat-treated glass.
The window cleaners' association pushed back, publicly and hard. In 2007 the IWCA's Tempered Glass Quality Committee formally asked GANA to retract two bulletins — the cleaning procedures one and Heat-Treated Glass Surfaces are Different — arguing they were misleading. The IWCA's position, still published today, is that scrapers have been standard practice for decades across the window cleaning, glass, and building trades, that millions of them have been sold, and that scrapers won't scratch uncoated glass when properly used — the implication being that if a pane scratches under a properly used blade, the pane was defective.
The two positions are not actually contradictory about the physics. They're both true. Debris-free uncoated glass tolerates a properly used blade; glass with fabricating debris doesn't. The fight is over which of those two facts gets written down as the default, because the default decides who eats the cost of a scratched curtain wall.
They have since converged enough to publish the cleaning bulletin jointly. That's a diplomatic achievement, and the underlying tension is unchanged.
Our position, for whatever an outsider's is worth: we don't sell glass and we don't clean windows for a living, so we have nothing to protect here. The technically honest reading is that fabricating debris is a manufacturing quality problem, that it is the actual cause of most tempered-glass scraping scratches, and that this does not help you at all — because you cannot see it before you scrape, and the fabricator will not be there when you do. The asymmetry is what matters: the cost of assuming debris is present and being wrong is ten extra minutes. The cost of assuming it's absent and being wrong is the pane. Act accordingly, and don't mistake being right about the cause for being covered for the consequence. That's what the scratches piece is about at length.
Most people reach for the blade first. Try this first instead — on latex, it's frequently the whole job.
Water-based paint (latex, acrylic, most interior wall paint, most spatter): soak it. Hot water, a few drops of dish soap or the house-standard mix, applied with a cloth or a strip washer and left to dwell for five to ten minutes. Latex is water-borne and never fully crosslinks into anything a solvent-grade chemist would call cured; hot water swells the film and it releases from a non-porous surface like glass surprisingly willingly. Re-wet it as it dries. Then push at an edge with a plastic scraper, a fingernail, or the corner of a credit card.
A large share of latex spatter, even months old, comes off exactly like this. If it does, you're finished, and you never touched a blade.
Oil-based and alkyd paint doesn't cooperate with water, because it cured by oxidation into a genuine polymer film. You're on the solvent path — mineral spirits first, and the solvent ladder covers the escalation properly. Wet a cloth, hold it against the spot, let it dwell, and expect the softened film to come off with pressure and patience rather than immediately.
Spray overspray is a special case worth understanding: it lands as a mist of tiny droplets that flash off almost instantly, so it's thin, weakly bonded, and covers a large area. Solvent-wet the whole pane, let it dwell, and a large share of it wipes off with a cloth alone. This is the one case where the area looks alarming and the bond is trivial.
Cured two-part coatings — epoxy, urethane, powder-coat overspray, anything that came out of a commercial gun: stop. These have crosslinked into something that will outlast the blade, and this is a restoration job. Call somebody.
Paint isn't usually alone. The same job normally carries three relatives, and they behave differently enough to be worth separating.
Adhesive residue from stickers and labels is not paint and does not want a blade first. It wants heat and time. A hairdryer at moderate heat for thirty seconds softens the adhesive enough that the label lifts in one piece rather than shredding — and shredding is what turns a two-minute job into twenty. Whatever tacky film remains comes off with a citrus-based adhesive remover or mineral spirits on a cloth, dwelt for a minute. Reach for the blade only for a stubborn edge, and only after the three questions.
The manufacturer's own paper label — the sticker on new glass giving the model and the ratings — is the special one, because it has been baked onto the glass by however many months of sun it sat in before anybody peeled it. Soak it thoroughly, let the water penetrate the paper, and expect to soak it twice. Do not dry-scrape a sun-cured label; that's the identical geometry to dry-scraping paint, and this one has grit and paper fibre in it.
Construction dust cemented on by rain is the sleeper. It looks like nothing and it's the thing most likely to scratch, because it's mineral, it's hard, and it's sitting exactly where you're about to run a blade. The trade's construction-site protection bulletin exists mainly because of this, and the scratches piece covers the aftermath: glass on a job site collects cement slurry, mortar splash, cut-tile mist and drywall dust, and once that dries in place a blade doesn't cut it — it drags it. Flood the pane and rinse it before any blade touches it, every time. Much of what gets recorded as scraping damage is really a razor pushing a grain of construction sand.
If you've cleared the three questions and the boring method didn't work:
New blade. One inch. Not a 4-inch, not the one in the drawer that scraped tile last summer. A used blade has a rolled or nicked edge and it drags material; the width matters because a wide blade cannot be kept in full even contact by hand, and its corners lift and dig. The bulletins are specific about the 1-inch size for a reason.
Flood the glass first, and keep it wet. This is the single most important line here. Dry scraping is how the overwhelming majority of scratch damage happens, because water is what floats loose grit up and away from under the edge. Wet the pane, scrape wet, re-wet constantly. If the blade squeaks, it's dry — stop and wet it.
Low angle, around 30 degrees, one direction only. Push away from you, lift, return, push again. Never back-and-forth. A returning blade drags whatever it just collected back across the glass with the edge on top of it, which is the exact geometry for trapping a particle and drawing a line with it. The bulletin's warning about back-and-forth motion exists because of that mechanism.
Spots only, not the pane. You are removing paint from where paint is, not resurfacing a window.
Test first, somewhere invisible. A bottom corner, behind where the blind stack sits, the edge that the frame covers. Scrape two inches. Dry it. Look at it in raking light from a low angle — scratches from this hide in normal light and appear when the sun is low, which is why they're always discovered a week later. If that test patch scratches, stop entirely. You've found fabricating debris or a coating, you've done it in a place nobody will see, and that test just paid for itself several hundred times over.
Listen. Clean glass under a wet blade is silent and smooth. Grit announces itself as a gritty tick you can feel through your fingers before you can see what it did. The moment you feel it, stop.
Everything above is technique. This part is the business, and it's the part that ends companies.
The scratched-glass dispute has a completely predictable shape. The builder says the window cleaner scratched the glass. The window cleaner says the glass arrived defective. The fabricator says its glass met spec. Everyone is partly right, nobody can prove their part after the fact, and the cheapest party to blame is the one with the smallest legal budget — which is the window cleaner, every time.
The IWCA's own guidance to members is unromantic about this: have the builder sign a tempered-glass scratch waiver, and educate builders about glass quality problems before they become scratches. The association publishes a sample waiver and then tells members to have their own lawyer look at it, which is the correct advice and also a fairly clear signal about how often this goes to a lawyer.
The conversation to have, before the blade comes out: this glass may carry fabricating debris; debris is a fabrication defect and not a cleaning defect; here is the test patch and here is what it showed; if you want the paint off tempered glass of unknown provenance, this is the risk and here's who carries it. In writing. Before.
Because the phone call afterward is not a negotiation. It's an invoice.
Not steel wool, not a scouring pad, not "fine" abrasive anything. They work, and they leave a haze of fine scratches that shows up in low sun forever. The fact that the paint left is not evidence the glass survived.
Not acetone near the frame. On glass it's fine and it evaporates clean. On vinyl sash, painted frames, gaskets and glazing beads it's a solvent for the frame too, and it will bloom or soften them permanently in seconds. Same caution as any strong solvent — the substrate around the glass is almost always the more fragile material.
Not a pressure washer, which won't move cured paint anyway and will drive water past the weatherstripping.
Not a blade on a hot pane in direct sun. The bulletin advises against cleaning glass in direct sunlight generally — everything flashes off before it can dwell, and you end up scraping something half-dry. Work the shaded elevation, or work early.
And not the blade at all if you haven't answered the three questions. Which is where we came in.
Tony is a regional contributor covering the Mid-Atlantic from Falls Church, with a route built on Tysons-Reston coated-glass commercial work and the post-construction cleans that come with it. Articles bylined by Tony draw on that fieldwork. He has been on both sides of a scratched-glass dispute and would rather nobody else had to be.