Google Ads Product Titles Report (August 2026): What Small Ecommerce Teams Should Do When AI Rewrites the Title They Wrote

On August 26, 2026, a new Product titles report started showing up inside Google Ads accounts. It lists a sample of the titles Google’s AI wrote and served in your Shopping ads in place of the ones you uploaded, with your original sitting in the next column so you can see what got replaced.

For a large retailer this is a reporting curiosity. For a small ecommerce brand the product title is close to the entire pitch — the only sentence you control in a grid comparing you against nineteen other thumbnails at the same price. Here are the two titles head to head, and a rule for which one to fight for.

What is the Google Ads Product titles report?

A beta report showing a sample of AI-generated product titles served in your Shopping ads, next to the original from your feed, with performance data for each. It does not let you approve or reject titles — a window, not a switch.

It was first spotted by performance marketer Yash Mandlesha on LinkedIn, then written up on August 26 by Barry Schwartz at Search Engine Roundtable and Anu Adegbola at Search Engine Land. The columns:

• The customized product title created by Google AI
• Your original product title
• Impressions
• Product clicks
• CTR
• Cost
• Avg. CPC

Two caveats are printed on the report itself. It shows a sample. And AI titles only appear “once they’ve received enough impressions,” so low-volume SKUs — most of a small catalog — may be getting rewritten with no row to prove it. Google’s copy also says your titles “will still be shown in ads when they’re the most relevant,” making this a per-auction swap, not a permanent replacement. It differs from the product title A/B experiments report earlier in 2026: that one tested titles you chose.

Why is Google rewriting product titles at all?

Because it optimizes for query relevance, and query relevance and differentiation are different jobs. Google substitutes its title when it predicts the generated version matches the query more closely than yours.

Read that through positioning — the Ries and Trout school, where a brand’s job is to occupy a distinct slot relative to the alternatives beside it. A Shopping grid is the purest positioning environment in paid media: twenty near-identical rectangles, each with an image, a price, and one line of text. That line is your whole position.

The mechanism, which predicts cases this article doesn’t cover: Google scores relevance vertically, your title against one query in isolation. Differentiation is scored horizontally, against the nineteen competitors rendered beside you. The auction only sees the first. So a rewrite systematically trades distinguishing attributes for query-term coverage — the only variable it is graded on.

The prediction that falls out: the more of your margin depends on an attribute nobody types into the search box, the more a rewrite costs you. “Handmade in Vermont,” “set of 4,” “ships in 24 hours,” “fits 2019–2024 models” — rarely searched, frequently the reason someone picks you over the cheaper listing. Flipped, the same rule says when the rewrite helps: if you resell a commodity SKU where the query is the product, Google’s version will probably beat yours, and you should let it.

Does an AI-rewritten product title actually beat the one you wrote?

The report cannot tell you, and that is the most important thing about it. Every column — impressions, clicks, CTR, cost, avg. CPC — measures the top of the funnel. None measures conversion rate, revenue, average order value, or returns.

I don’t read that as Google hiding something; product-level conversion attribution in Shopping is messy, and shipping the clean metrics is a reasonable call. But notice the result: the scoreboard is built entirely out of numbers a broader, more query-matched title is almost guaranteed to improve. A change that widens reach wins every column shown, whether or not it wins in your bank account.

Run it with real figures. A home-goods seller spends $1,800 a month at a $0.62 average CPC: roughly 2,900 clicks, 1.8% conversion, 52 orders, $74 average order value. Google’s rewrite lifts CTR 11% and pulls CPC to $0.58, so the same $1,800 buys about 3,100 clicks. That row is green in all five columns.

Except the rewrite dropped “Set of 4.” Some of those extra clicks are people expecting one item at that price, and conversion slips to 1.5% — 47 orders instead of 52, about $370 less revenue on identical spend, plus the support email. Nothing in the report shows you that, and on a $1,800 budget $370 is the difference between a channel you scale and one you quietly switch off in November.

How should a small team read this report without losing a day to it?

Treat it as a quarterly feed audit, not a daily monitoring task. Seven steps, one afternoon:

1. Pull the report and sort by cost, descending. Your catalog might have 400 SKUs; 20 carry the spend. Only those 20 rows matter this quarter.
2. Diff the two titles by hand. Write down only what the AI version removed. Additions are harmless; deletions are where the money is.
3. Sort every deletion into two buckets: noise (a repeated brand name, a redundant colour word) or a purchase reason (quantity, compatibility, material, warranty, shipping speed). Ignore the noise.
4. Move every deleted purchase reason into a structured feed attribute — size, material, product_highlight, shipping_label. Attributes are harder for a rewrite to discard than adjectives buried in a 140-character string.
5. Front-load the purchase reason in your own title. If “Set of 4” is doing the selling, it belongs in the first 35 characters — mobile Shopping truncates, and so, in effect, does the rewrite.
6. Add text guidelines in campaign settings naming the terms that must survive.
7. Set one reminder for 30 days out and compare product-level conversion rate, not CTR, for those 20 SKUs. Less than 30 days and a small account’s numbers mean nothing.

Write the outcome down. A one-page title policy — which attributes are non-negotiable, which are decoration — outlasts any week of report data and survives the next freelancer who touches the feed. Keeping it with your campaign notes at campaignpress.ai beats keeping it in your head.

What can you actually control, and what is the text guidelines setting?

One lever, and it is indirect. The report’s help text points to text guidelines in your campaign settings: “You can help Google AI to create on-brand ad copy by adding text guidelines in your campaign settings.”

Read the rest of that note. Google adds that if you changed your guidelines during the selected date range, “you may still see ad copy that violates your guidelines, but it wouldn’t show in your ads today.” Translation: guidelines are a prompt, not a filter, and the report is backward-looking. Don’t judge a guideline change on data from the week you made it.

The stronger lever is upstream. Feed attributes and a title whose first 35 characters already say the thing you’d fight to keep shape what a rewrite has to work with. You cannot veto the AI title — you can make your own the one it has least reason to replace.

Which September and October dates interact with this?

Feed work competes with everything else on the Q4 runway.

• September 1–30, 2026 — Google Ads AI Max auto-migration window. If Search shifts the same month you change titles, neither result is attributable. Feed audit first.
• September 4, 2026 — deadline to opt out of Performance Max video enhancements. PMax carries your Shopping inventory, so it touches the same campaigns.
• September 30, 2026 — Google’s updated alcohol advertising policy takes effect, per Search Engine Roundtable’s August 28 report.
• October 1, 2026 — Microsoft Ads removes Max CPC bidding; Local Services Ads begins charging for qualifying missed and follow-up calls. Neither is Shopping, both eat the same team’s October.
• Mid-October onward — Shopping CPCs climb into Black Friday. A title change made in November is one you cannot read, because the auction moved underneath it.

Do this in September. In November you are buying, not testing.

Frequently asked questions

Can I turn off AI-generated product titles entirely?

There is no documented opt-out toggle as of August 29, 2026. The only stated control is text guidelines in campaign settings, which steer generation rather than disable it. If a hard opt-out ships later, expect it at campaign level rather than per-SKU, since that is how Google has shipped comparable asset controls.

My report is empty. Does that mean Google isn’t rewriting my titles?

No. The report is in beta, hasn’t reached every account, and only populates rows once an AI title has “received enough impressions.” A catalog of low-volume SKUs on a modest budget can be getting rewritten steadily and still show nothing. Absence of rows is not evidence of absence of rewrites.

Should I just copy the AI title into my feed if it performs better?

Only if it beat yours on conversions, which the report does not tell you. Check product-level conversion rate over at least 30 days first. And if you adopt it, you have handed your differentiation to a system that keeps optimizing toward the category average — fine for commodity resale, bad for a brand with a real product story.

What if I only have 20 products and no time?

Then it is a 40-minute job. Diff the titles for your top five sellers, move any deleted purchase reason into a structured feed attribute, and stop. Small catalogs concentrate spend so heavily that five SKUs capture most of the exposure. Teams sizing tooling against media budget can weigh that against what a plan covers.

So which title should you defend?

If the query already contains everything that makes your product worth buying, let Google write the title — your position was never in the words. If the reason you win is an attribute nobody searches for, defend it: front-load it, structure it, name it in your text guidelines, and check conversion rate rather than CTR. The report beats the previous state, which was not knowing. Just remember what it counts, and what it leaves to you.