The rule before the formula
A Google Shopping title should identify the exact product and variant a shopper reaches after the click. It is product data, not a miniature search ad. Useful titles put distinguishing facts in a deliberate order; weak titles copy a poetic website name or repeat keywords until the field is full.
Google’s title and structured-title specification makes the title required for each product. The standard title value is 1–150 characters. Google may truncate titles to fit, so 150 characters is a maximum, not a target.
Requirements and recommendations
| Requirement | Recommendation |
|---|---|
| Describe the product accurately and match the landing page | Put the product type and most decisive attributes early |
| Distinguish variants and use the correct language | Use a category-specific attribute order |
| Do not use promotional text, gimmicky capitals or foreign characters for attention | Write for a person scanning a truncated title |
| Stay within 1–150 characters | Remove facts that do not change product choice |
If generative AI created the title, Google directs merchants to use structured_title and identify the digital source type. That is a data-specification requirement, not a reason to mass-generate titles without checking the product record.
Build a category title map
List the attributes a buyer uses to distinguish products in each category, then rank them. Brand belongs early when people search for it or when it establishes compatibility; it can be later for unbranded decorative products. Variant attributes such as colour and size must describe the actual submitted variant.
| Category | Suggested order |
|---|---|
| Apparel | Brand + gender/audience + product type + material/style + colour + size |
| Footwear | Brand + model + gender + shoe type + key feature + colour + size |
| Electronics | Brand + model + product type + core specification + capacity/colour |
| Furniture | Brand/collection + product type + material + size + colour/finish |
| Consumables | Brand + product + variant/flavour + pack count + unit size |
| Compatible parts | Brand + part type + compatible model + part number + key specification |
These are recommendations, not mandatory Google formulas. Validate them against real query language and the product facts you can support.
Worked title examples
Apparel
Weak: “The Weekend Essential”. The shopper and Google cannot tell whether this is a shirt, bag or candle.
Better: “Ridge & Loom Men’s Merino Wool Quarter-Zip Jumper, Charcoal, Large”. The product type, material, colour and exact variant are explicit.
Electronics
Weak: “Wireless Headphones Bluetooth Headphones Best Noise Cancelling Headphones”. This repeats terms and omits the model.
Better: “Auralis H7 Wireless Noise-Cancelling Headphones, 40-Hour Battery, Black”. Keep the battery claim only if it is accurate on the page.
Furniture
Weak: “Oxford Table”.
Better: “Hartwell Oxford Oak Bedside Table, 3 Drawer, Natural Finish, 62 cm”. Dimensions can be highly useful when they are a buying constraint.
Compatible product
Weak: “Premium Replacement Filter”.
Better: “PureFlow Replacement HEPA Filter for AirClean AC300, 2 Pack”. Do not imply the product is made by AirClean if PureFlow is a third-party brand.
Need a title system for a large catalogue?
Upscale can map query language to reliable product attributes, deploy by cohort and measure the result at item level.
Book My Free Ad AuditHandle variants correctly
Every sellable variant needs a unique ID. Related variants should share an item_group_id where the specification requires grouping, and variant attributes such as colour, size, pattern or material should describe the selected item. The landing page must preselect or clearly present that same variant.
Do not submit a parent title saying “Available in 12 Colours” while the item’s image and price represent one blue variant. Use the blue variant’s title and colour. If sizes share one landing page, make sure the submitted size is selectable and available at the advertised price.
Use real query data without keyword stuffing
Collect category-language evidence from Google Ads search-term insights, onsite search, organic queries, customer-service language and competitor-result review. Map the language to true product attributes. If shoppers use “sofa bed” and the catalogue calls it a “convertible sleeper”, the title can include the common product type provided it accurately describes the item.
Frequency is not permission to add an irrelevant term. “Cheap”, “best”, “free delivery” and sale messaging belong in pricing, promotions and shipping settings, not in the title. Model numbers and compatibility deserve priority when they define the intended product.
Prioritise the catalogue
Begin with products that have meaningful impressions and below-category CTR, high spend with questionable query relevance, or high commercial potential but vague source titles. Separate title problems from offer problems: a clear title will not fix an uncompetitive price or weak image.
For large catalogues, write transformation rules from structured attributes rather than editing strings by hand. Add fallbacks for missing fields and prevent duplicated words. Review a sample from every category, plus the longest and shortest generated titles, before publishing.
Measure a title change
Create a fixed cohort of item IDs and a comparable holdout if volume permits. Record the change date, old title and new title. Keep price, bids, imagery and promotion stable where possible. Over a period long enough for normal conversion lag, compare impressions, CTR, search relevance, qualified-session rate, conversion rate and contribution.
Read the sequence. More impressions with much lower post-click quality may mean the title broadened matching without improving intent. Higher CTR with unchanged conversion rate may still be useful, but only if incremental clicks are profitable. A small catalogue may need a before/after analysis rather than a statistically clean experiment; state that limitation.
Pre-publish title QA
- The exact product type is obvious in the opening words.
- Brand and model are accurate; compatibility is not misleading.
- Colour, size and other variant facts match image and landing page.
- No promotion, shipping claim, excessive capitals or keyword repetition.
- Length is within 150 characters and the useful information is front-loaded.
- The rule handles missing attributes without stray punctuation or duplicates.
Turn source attributes into a safe title rule
A title template is only as reliable as its inputs. Create a field dictionary for brand, product type, model, material, feature, colour and size. Specify fallback behaviour when a value is absent. Do not insert “N/A”, duplicate the brand because it also appears in the product name, or leave double separators when a field is blank.
A simple transformation for furniture might be:
- Normalise brand and collection names.
- Choose the most specific product type from a controlled map.
- Add material only when it describes the submitted product.
- Add one decision-making dimension or size label.
- Add colour or finish for the exact variant.
- Remove duplicate tokens and trim safely below 150 characters.
When trimming is necessary, remove lower-priority attributes rather than cutting a word or model number. Keep the original source values and generated title in a QA export.
Additional category examples
| Source name | Improved hypothetical title | Why |
|---|---|---|
| “Classic No. 4” | Field & Found Classic No. 4 Leather Crossbody Bag, Tan | Adds product type, material and variant colour |
| “Ultra 10” | Vela Ultra 10 Android Tablet, 10.1-inch, 128 GB, Wi-Fi, Grey | Identifies model and decisive specifications |
| “Daily Restore” | Northwell Daily Restore Moisturiser with Ceramides, Fragrance Free, 50 ml | Adds product form, feature and volume without claims |
| “AC300 Filter” | PureFlow HEPA Filter Compatible with AirClean AC300, 2 Pack | Clarifies third-party compatibility and pack count |
For cosmetics and health-related products, avoid unsupported therapeutic claims. For compatibility products, make the third-party brand clear and do not imply manufacturer endorsement.
Failure cases in automated title generation
- A colour field contains “Assorted”, while the image shows one specific colour.
- The model appears twice because the source product name already includes it.
- Size values lose units, turning 50 cm into 50.
- Parent and child attributes conflict, such as “Blue” in the source name and “Red” in the variant field.
- Promotional tags from the ecommerce title, such as “Sale” or “Free Gift”, pass into the feed.
- Every title changes on each run because attribute order is unstable.
Block publication when the QA report finds contradictory variant values, missing product type in a priority category or unexpected mass changes. Correct the source or rule, regenerate and resample.
A more rigorous title test
Match test and holdout products on category, recent impressions, price band and performance where possible. Freeze membership before release. Track processed-title acceptance and product status before measuring media. Use at least one full business cycle and normal conversion lag; seasonal catalogues may require a different design.
Report impressions, clicks, CTR, qualified sessions, add-to-baskets, purchases and contribution for both cohorts. If the test gains 20% impressions but loses 15% conversion rate, calculate absolute orders and contribution rather than declaring victory on visibility. Preserve search-relevance evidence to understand what changed.
The feed optimisation checklist covers release QA around title deployments.
Maintain a category-level title specification with examples of accepted and rejected outputs. Review it when product attributes or search language change. A rule that worked for last year’s catalogue may omit a new technical specification or generate misleading combinations for a new product family.
For multilingual feeds, build and review titles in the target language rather than mechanically appending translated keywords. Product facts, units, variant values and landing-page language must remain consistent for the target country.