Guide
AI product title generator for feeds
Most 'AI title generators' invent names for a new product. A feed needs something else: a listing title that tells Google, Microsoft or Meta exactly what the product is, built from the brand, type, model and attributes you already have. Here is what good AI output looks like for four kinds of catalog, when a plain rule does the job better, and how Textfinity generates, reviews and publishes AI titles for every product.
We connect your store, build the feed and keep it in sync. Updated 2026-10-27.
At a glance
- A listing title is not a product name: it is product type, brand, model and the attribute that separates the variant, inside Google's 150 characters with the deciding words in the first 70.
- Use a rule (combine, advanced) when the attributes exist as columns; use AI when the facts sit in the description or the store title is a bare name.
- Textfinity's AI configs write a 100 to 150 character title per product from the fields chosen, front-loaded into the first 70, optionally translated, and only accepted titles reach the feed.
- We set up the AI configs and the rule-based titles in Textfinity for you; pricing is quoted per catalog after a short call.
What an AI product title generator should produce for a feed
A Shopping title does the job a keyword list does in search ads: it decides which queries the product can show for. Google's own guidance is to use the product name, brand and specific details shoppers search for, to include variant details such as color and size, and to keep promotional text, capitals for emphasis and symbols out. Brand naming tools optimize for memorable; a feed title optimizes for matched and clicked.
So judge an AI product title generator on three things. Does every word come from a fact you supplied, not invented? Does the product type appear in the first 70 characters, where shoppers see it? And can it do this for the next 5,000 products with the same structure, so your titles look like one catalog rather than 5,000 separate writing sessions?
Rule-based titles or AI titles: which products need which
| Situation | Better tool | Why |
|---|---|---|
| Brand, type, color and size are clean columns | Rule: combine or advanced | Deterministic, instant, no review needed |
| Store titles carry promo text or stray capitals | Rule: remove from regexp | A pattern fixes all of them at once |
| Store title is a bare name ('The Willow') | AI config | The product type has to be read from the description or category |
| Specs are buried in a long description | AI config | A model picks the attributes that matter per product |
| Titles needed in another language | AI config with translation | Translated and restructured in one pass |
| A channel wants a 65-character title | Rule | AI titles aim for 100 to 150 characters; build a short combine instead |
How Textfinity generates AI titles for a whole catalog
An AI config reads the source fields chosen for it (title, description, category, manufacturer, and any extra attributes as additional context) and runs Enhance Title on every product, or on a filtered subset such as titles shorter than 40 characters. The generated titles target 100 to 150 characters with the key information in the first 70. With a target language (German, French, Dutch, Swedish, Italian, Polish and most other European languages; not Spanish) the title comes out in that language.
Jobs run in batches of up to 50 products. Results go into a review queue where they are accepted, edited or accepted all at once. In the feed, replace from AI config uses accepted titles and falls back to a chosen field; a conditional rule can instead fall back to a rule-built title. New products and changed products are generated again, and editing the config creates a new revision. Processing runs on EU-hosted infrastructure.
Google's rule for titles written by generative AI
Google's title specification asks that titles created with generative AI be submitted in the structured_title attribute, with the digital source type trained_algorithmic_media. In a text feed the value looks like trained_algorithmic_media:"Northpeak Trail Runner X Men's Waterproof Shoes". If both title and structured_title are sent, Google uses title.
Textfinity's replace from AI config writes the generated text into whichever column it is mapped to, as plain text; it does not add the trained_algorithmic_media prefix. Read Google's current guidance for your account before sending AI titles at scale, and keep the review step: it is also how you make sure the title still matches the landing page.
Examples
Worked examples, by situation
Four catalogs where a rule alone was not enough. Every word in each 'after' comes from a field listed in the 'before'.
Model name into a searchable title
Before
title: Trail Runner X manufacturer: Northpeak · category: Men > Shoes > Trail Running description: Waterproof Gore-Tex lining, Vibram sole. Grey/orange.
After
Northpeak Trail Runner X Men's Waterproof Trail Running Shoes, Gore-Tex Lining, Vibram Sole, Grey/Orange
How we set it up
We set up an AI config with Enhance Title; source fields title, description, category and manufacturer; filter: title is shorter than 30. title uses replace from AI config, fallback title.
A Dutch title from English attributes
Before
title: Ripple Vase description: Hand-glazed stoneware vase, 25 cm, matte white. vendor: Hollow & Co
After
Hollow & Co Ripple vaas van handgeglazuurd steengoed, 25 cm, mat wit
How we set it up
We set up an AI config: Enhance Title with translation Dutch, source fields title, description and manufacturer (Shopify vendor). The NL feed maps title with replace from AI config, with the English title as fallback until accepted.
Specs pulled out of a long description
Before
title: USB-C Cable 2m manufacturer: Cablecore description: USB-C to USB-C, supports 100W PD charging and 10Gbps (USB 3.2 Gen 2). Braided nylon jacket, black...
After
Cablecore USB-C to USB-C Cable, 2 m, 100 W Power Delivery, USB 3.2 Gen 2 10 Gbps, Braided Nylon, Black
How we set it up
We set up an AI config with description as a source field so the model can read wattage and data rate. The supplier XML is a Google-shaped RSS feed; other XML schemas need a CSV export first.
AI title for Google, rule-built short title for Meta
Google feed: title mapping (conditional)
if [id] is accepted in AI config "Titles EN"
→ replace from AI config "Titles EN"
else
→ combine [Vendor], [Title], [metal]
delimiter " "How we set it up
We give the Google feed a conditional title rule: the accepted AI title, and a rule-built title for everything not yet reviewed. The Meta feed uses only the combine rule, which stays near Meta's recommended 65 characters, since AI titles aim for 100 to 150.
Copy this plan
Plan: AI titles for the products that need them
Rules first, AI for what is left. That keeps the review queue short.
01
Send us the catalog and profile the titles
Check the title length distribution and which attribute columns are filled. Empty brand or type columns point to AI; full ones point to rules.
What we set up: We connect the store and read feed hygiene's length distribution and field completeness with you.02
Build the rule-based title first
Combine brand, title, color and size for the Google template. Every product gets a decent title before any AI runs.
What we set up: We build the Google Shopping feed with title as a combine or an advanced rule.03
Point AI at the weak titles
Decide which source fields the model reads (include the description and category), whether it needs a target language, and which products the rule cannot help.
What we set up: We set up the AI config: source product data, Enhance Title, translation and a filter.04
Review and accept
Read a sample per category, edit the misses, then accept. Unaccepted titles never reach the feed.
What we set up: We go through the pending generations with you and accept them, one by one or with Accept all.05
Switch the feed to AI titles with a rule fallback
A conditional mapping uses the AI title when accepted and the rule title otherwise.
What we set up: We give title the condition is accepted in AI config → replace from AI config; else → combine.06
Publish and measure
Merchant Center fetches the feed; compare clicks for AI-titled and rule-titled products in Google Ads after a few weeks, outside Textfinity.
What we set up: We register the feed URL in Merchant Center with you: you give us access or we walk you through it.
Customization tips
Make it yours
Model numbers belong in the input
Shoppers search exact model numbers for electronics and spare parts. Pass the MPN or model column as additional context; a model cannot put in a title what it never saw.
Do not AI-title products a rule already handles
If brand, type, color and size are clean columns, a combine rule gives the same title every time with no review. Spend AI and review time on the bare names and messy supplier data.
Expect regeneration when you change the config
Every config edit makes a new revision and regenerates the products, which means a new review round. Settle the source fields on a small filtered subset before running the whole catalog.
Keep the landing page in step
An AI title that names a color or size the link does not open is a mismatch Google can disapprove. Variant-level ids and links make this safe; a product-level link with a variant-level title does not.
FAQ
Questions about ai product title generator
Is an AI product title generator the same as a product name generator?
No. A product name generator invents a brand or product name for something new. A product title generator for e-commerce writes the listing title shoppers see in Google Shopping or a marketplace, from facts about an existing product: type, brand, model, material, color and size.
Can I use AI-generated titles in Google Shopping?
Yes, provided they follow Google's title rules: up to 150 characters, no promotional text, no capitals for emphasis, consistent with the landing page. Google asks that titles created with generative AI be sent in structured_title with the digital source type trained_algorithmic_media.
How long should an AI-generated product title be?
For Google Shopping, up to 150 characters with the product type and key attribute in the first 70, which is what Textfinity's AI configs aim for. For Meta, which recommends 65 characters, a rule-built short title usually works better than trimming an AI one.
Is there a free AI product title generator in Textfinity?
No. AI titles run in AI configs that we set up on your catalog in Textfinity, priced per catalog after a short call. Rule-based titles with combine, regex and the advanced DSL fix most catalogs before AI is needed.
Set up for you
Have every weak title rewritten by AI, and reviewed before it ships.
Send us your catalog. We build the rule-based title first, then point an AI config at the products it cannot fix, with translation where you sell in another language.
Tell us your store and channels. We reply within 24 hours.