In a text ad you write headlines, you test them, and you can rewrite the pitch on a Tuesday afternoon. In a shopping ad you write almost nothing. What appears is your product image, your price and your product title, pulled from the feed.
That makes the title the most valuable string of text in the account, and it is usually the one written last, by whoever added the product, in whatever format the supplier sent.
It is doing two jobs at the same time. It is the main signal used to decide which searches your product is a candidate for, and it is the copy a shopper reads when deciding which of the products in front of them to click. Those two jobs pull in slightly different directions, which is what makes titles interesting to work on and why a spreadsheet full of supplier names is leaving money on the table.
One string, two jobs
The matching job wants the words a buyer would type. Category, brand, the attribute they would specify, the size or format they would name. Without those words present, the product is a weaker candidate for the search, no matter how good the image is.
The reading job wants a title that makes sense to a human in a crowded row of similar things. Someone scanning results is deciding in a second, comparing price against what they can tell from the picture and the first few words.
The conflict is that stuffing every possible keyword serves the first job and ruins the second. A title that reads like a list of search terms looks like the cheapest option in the row even when it is not.
The resolution is order rather than length. Put the words that must be matched at the front, in the sequence a person would say them, and let the specifics trail after. Titles are typically truncated on smaller surfaces, so the front of the string is doing most of the work in both jobs at once.
In shopping, your ad copy is your product title. There is not much else to write.
Front-load what people type
The single most common title fault is starting with something the buyer would never search: an internal SKU, a collection name, a season code, or the brand when nobody searches the brand.
Write the title the way a customer would ask for the product out loud. If they would say a mid century oak dining chair, the title should not start with the range name that only exists in your catalogue.
Brand placement depends on whether people search it. For a known brand, the brand at the front is what buyers scan for and it belongs there. For an unknown one, the brand at the front spends the most valuable words in the string on something nobody is looking for, and the product type should lead instead.
The practical way to find the right words is to read what people actually search to reach your products. Your own search terms report is the closest thing to a direct transcript of buyer language you will ever get. The phrases in it, particularly the ones that convert, are the phrases the titles should contain, in roughly the order they are typed.
A shape per category, not one template
A single title formula across a mixed catalogue produces awkward results, because different categories are searched with different attributes.
Apparel is searched by brand, gender, product type, colour, material and size, and shoppers filter hard on the ones that matter to them. A title missing colour or size on a variant level product is missing the words the buyer used.
Parts and technical products are searched by compatibility above everything. The model or fitment it works with is the search, and a title without it is invisible to the person who knows exactly what they need.
Consumables are searched by format and quantity. The number of units, the weight, the concentration. Someone comparing prices needs the quantity in the title or your price looks wrong next to a smaller pack.
Homewares are searched by dimension and material more than by name.
So write a shape per category, apply it as a rule in the feed rather than editing thousands of products by hand, and let the exceptions be exceptions. The rule should build the title from fields you already hold, so new products inherit it automatically.
The words that cost you money
Some things in a title actively work against you.
Internal codes and warehouse references consume the front of the string and mean nothing to anyone outside the business.
Promotional text belongs in the fields designed for it. A title containing a sale claim risks the product being rejected, and a rejected product earns nothing at all.
All capitals reads as shouting to a person and is often flagged. Emoji and decorative characters do the same.
Marketing adjectives are the quiet ones. Premium, luxury, ultimate, best. They take space at the front of a title that could hold an attribute someone actually searches, and they carry no matching value because nobody types them.
Duplicate words are the last category. A title that repeats the brand three times reads badly, and repetition does not improve the match.
The test to apply to any word in a title is simple. Would a buyer type it, or would a buyer need it to tell this product apart from the one next to it. If neither, it is taking a place from a word that would.
Test titles the way you would test ads
Titles are testable, and almost nobody tests them, largely because they live in a feed rather than an ads interface and feel like data instead of copy.
Run changes as a proper before and after. Take a group of products with enough impressions to be readable, change one thing about the title shape, and hold everything else steady. Give it long enough to gather real numbers, and compare against a group you did not touch so a change in the market does not get credited to the change you made.
Watch two things: how often the product appears at all, which tells you about matching, and click through rate, which tells you about reading. A change that raises impressions but lowers clicks has probably matched you into searches you should not be in, and that is not always a win.
Record the shape you tested and the outcome somewhere permanent, because titles get regenerated whenever the catalogue is rebuilt and the lesson is easy to lose. A tested shape written into the feed rules survives. A one time manual edit does not.
Where the rest of the feed takes over
The title cannot carry everything, and trying to make it do so is what produces the unreadable ones.
The structured attributes exist to hold the specifics: colour, size, material, gender, age group, pattern, condition, and the identifiers that connect your product to a known one. Populated properly, those fields do the filtering work so the title does not have to contain every possible qualifier.
The description holds the detail a careful shopper wants, and the image carries scale and quality faster than any words.
Think of it as a division of labour. Title for the search and the scan. Attributes for the filters and the match. Image for the decision. Price for the comparison. Landing page for the doubts.
When a feed is set up that way, title work becomes a clean, bounded job rather than an endless argument about what else to squeeze in. And it is a job worth doing early, because everything downstream, the bidding, the budget, the campaign structure, is competing for searches the title decided you were eligible for in the first place.