AI Product Name Generator for a Shopify Range: Shortlist and Checks

Ask the AI for 30 names in three distinct styles, cut them to five using written rules, then check each of the five yourself: trademark registers for every market you sell in, the .com and your preferred domain, your Shopify handle, and a plain search. The generator supplies ideas; it cannot tell you a name is free.

That last point is where most small brands trip. A chat assistant will sometimes say a name "appears to be unused" with complete confidence, often based on nothing more than its training text. Treat everything it produces as an unchecked suggestion. What follows is a worked example of the whole process, from a blank brief to a name you can print on a label, with prompts you can copy.

Follow me on Instagram@sagnikteaches

The example: a six-product candle range for a gift shop

Take an illustrative gift shop that sells online through Shopify and wants to launch its own range of six scented candles, made by a contract maker and labelled with the shop's own branding. The shop needs a range name, six product names, and confidence it is not about to step on someone else's mark. The owner has one afternoon.

Connect on LinkedInSagnik Bhattacharya

The whole process has five stages: write the brief, generate, filter, check, and record the decision. The first and third are where you add the value. The second is the quick part, and the fourth protects the investment.

Subscribe on YouTube@sagnikteaches

Stage one: a naming brief the tool can work from

Vague prompts produce the same dozen names everyone else gets: "Ember", "Haven", "Luna". Spend ten minutes writing a brief first. Cover:

• What it is and who buys it. "Hand-poured soy candles, $18 to $34, bought as gifts, customers aged 28 to 55."
• The feeling you want. Three words: "warm, a bit witty, not mystical".
• What to avoid. Words every competitor uses, anything that sounds like a medical claim, anything hard to spell when said aloud.
• Hard constraints. Maximum 14 characters, one or two words, pronounceable in English, no numbers.
• Names you like and why. Two or three examples from other trades, not competitors.

If you have a tone-of-voice document, paste it in; a brand voice guide the AI can follow is worth building once, because naming, descriptions and social posts all use it. For a shop with no document yet, three sentences describing how you would talk to a customer will do.

Stage two: ask for 30 names in three styles

Asking for a single list invites a pile of near-identical names. Splitting by style forces range. A workable prompt:

You are naming a new range of six scented candles for a gift shop's
own brand. Customers buy them as gifts. Tone: warm, a little witty,
not mystical or spiritual.

Give me 30 candidate names for the RANGE, in three groups of 10:
A. Real words used unexpectedly (e.g. a word from cooking or weather)
B. Short coined words, 2 to 3 syllables, easy to say aloud
C. Two-word phrases that sound like a place or a moment

Rules: maximum 14 characters including spaces. No numbers. No
medical or wellness claims. Avoid these overused words: ember, haven,
luna, bloom, glow, serene, zen. For each name give a 6-word reason.

A realistic slice of the output (illustrative) looks like this:

Do not trust the "reasons" as evidence of anything. They are the tool justifying its own choices after the fact. Use them only as a quick prompt for your own reaction. If you want more of one style, say so and ask for 15 more in that group, giving the three names you liked as a steer.

Stage three: filter 30 down to five with written rules

Do the filtering on a sheet with one row per name and one column per rule. Pass or fail each. A name that fails any rule goes, however much you like it. The rules for the candle range:

1. Say it on the phone test. Read it aloud to someone who cannot see it. If they spell it wrong twice, it fails.
2. Two-word search test. Type the name plus "candle" into a search engine. If the first page is full of other candles, shops or a famous brand, it fails.
3. Meaning check. Search the name plus "meaning" and, if you sell abroad, ask the AI what it means in the languages of your three biggest markets and then confirm with a native speaker or a reliable dictionary. A harmless word in English can be an insult elsewhere.
4. Length and label fit. It must fit on the label and in the Shopify product title without being cut off in search results.
5. Family fit. It must work as a range name that six product names can sit under.

In the illustrative sheet, "Simmer" fails the search test (several cookware and candle brands already use it). "Drizzle" passes. "Lumo" fails because the first page is dominated by an existing brand. "Wicka" passes, as does "Sunday Table". "Last Biscuit" passes but fails the family-fit rule, because the six scents are not food-themed. After filtering, five names remain: "Drizzle", "Wicka", "Sunday Table", and two others from the later batches. This is the list that goes forward to checking.

A tip on the sheet: record why each name failed. After a few launches you will see your own pattern, and it feeds the next brief. If most of your discards are for being too common, add that to the brief as a constraint.

Stage four: check each of the five, in this order

Check one: search the trademark registers yourself

Trademarks are registered per country or region, so the register to search depends on where you sell. WIPO's Global Brand Database brings together international registrations under the Madrid System and records from many participating national and regional offices, and TMview is a freely accessible search that pools records from dozens of national and regional offices. Each country's own office also has a public search. Use the register for every market where you plan to sell, not only the one where you are based.

How to search usefully:

• Search the exact name, then spelling variants and sound-alikes ("Wicka", "Wicker", "Wika").
• Look at the goods and services classes. Candles sit in one class, and clothing or software sit elsewhere. A match in a very different class is often lower risk but is not automatically safe, and famous marks can have wider protection.
• Note the status of each hit: live, pending, expired or abandoned. An expired mark is a different situation from a live one.

For an illustration, say "Wicka" returns one live mark for scented candles in a market you plan to sell in. That is a stop for that market. A near match in an unrelated class, say for software, is a note for an adviser rather than an automatic no. Whatever you find, the register search is a screening step, not legal clearance. If you plan to apply for your own trademark, or the name matters to the business, pay a trademark specialist or similar qualified adviser to run a proper clearance search before you commit.

Check two: domains and social handles

Look up the exact name as a domain with your preferred ending, then reasonable variants. A registrar's search box or a domain lookup tool will show whether it is taken and whether the site is live. A taken domain with an unused holding page is different from one with an active shop selling the same goods. Also check the same name on the social platforms you use, because a mismatch between your store name and your handle costs customers who go looking.

If a name is excellent but the plain .com is taken by an active business in a similar trade, drop it. If it is taken by something unrelated, you may still use it with a descriptor, such as the name plus "candles" or "home", but be honest about how many people will end up on the other site by mistake.

Check three: the Shopify handle and your own store

In Shopify, a product or collection gets a "handle", which is the last part of its web address. Shopify's developer documentation explains that the handle is generated from the title when the item is created, and that a duplicate title gets a number added to the end, so a second "Sunday Table" collection would become sunday-table-1. Create the draft collection with the range name and see what the address looks like. Check it is short, readable and not already used by another collection on your store.

The same documentation notes that changing a title later does not change the handle. If you are renaming existing products, edit the handle yourself and then create a URL redirect from the old address, which Shopify's help pages show how to do under the store's URL redirects, so you do not lose visitors or search standing. An illustrative slip: a shop retitles a candle from "Fig Candle" to "Second Helping", leaves the handle as fig-candle, and later wonders why the address no longer matches the name on the label.

Check four: a plain search and a marketplace sweep

Search the name on the big marketplaces where candles and gifts are sold, plus an image search for logos. Ten minutes here finds conflicts the registers miss, for example an established unregistered brand with a loyal following. Unregistered use can still cause legal problems in many places, so it counts.

Stage five: name the six products, then record the decision

With the range name chosen, say "Sunday Table", the six product names follow from the scents. Ask the AI for names that sit under the range, in a table with the scent notes you supply, and repeat the same filter. An illustrative result:

Run each product name through the same search test, although the trademark stakes are lower for product names inside a range than for the range name itself. "Boxing Day" is a common phrase and would be hard to protect on its own, which is fine for a product name but worth knowing. The test is whether another candle brand already sells a product with that exact name in your market.

When the names are final, write them on one page: the name, the date you checked, the registers you searched, what you found, the domain status and who decided. That page is cheap and gets used later, for example when a competitor writes to you or when you take on a new market. If a conflict appears in a year, you can show you made a reasonable effort. It does not remove the risk, but it shows what you knew.

Realistic mistakes with AI naming

Trusting the tool's "this name is unique" claim

In an illustrative case, the AI described a coined name as "unique and unclaimed". A two-minute register search showed an existing mark for home fragrance. The assistant had not searched any register and was offering reassurance, not information. Add a line to your prompts: "Do not claim any name is available or unique; I will check." It will still sometimes say it. Ignore it.

Falling for a name because the explanation is clever

Three of the 30 names had good-sounding stories attached. None of the stories would appear on a label. A name has to work with no explanation in the second it takes someone to see it on a shelf or a thumbnail. Hide the reasons column when you filter.

Skipping the sound-alike search

Trademark disputes are often about names that sound or look similar, not names that are identical. A search for only the exact spelling misses these. Search variants, especially where the vowels change.

Choosing a name that cannot be shortened

The Shopify title, the social handle and the label all have different space limits. A 14-character name fits the label, but with the range name and size it may be cut off in a search result. Build the full title early and look at it in the preview.

Naming with no brief, then regenerating forever

If you run the same loose prompt five times and get 150 names, you have a worse position than with one tight brief and 30 names, because the filter then consumes an evening. Fix the brief, not the volume.

A weak brief against a tight one, side by side

To see how much the brief matters, compare two prompts for a different illustrative shop: a pet shop launching its own range of dog treats. The weak prompt reads "Give me names for dog treats." It returns a list heavy with "Paws", "Bark", "Waggy" and "Tail" in various combinations, almost all of which will collide with existing brands and with each other.

The tight version names the buyer (owners who check ingredient lists), the feeling (honest, a bit cheeky), the product facts (single-ingredient, baked, 150 g bags), the banned words (paw, bark, wag, tail, woof, good boy) and the format (three groups of ten, maximum 12 characters, with the reason in six words). The returned list has far fewer clichés. In an illustrative run, perhaps six of thirty survive the search test instead of two of thirty. The extra five minutes on the brief saves the evening you would have spent sifting.

What the work costs in time and money

A quick sum for the candle range, with illustrative numbers. Brief: 20 minutes. Generation and one refinement round: 15 minutes. Filtering 30 names on the sheet: 40 minutes. Checks on five names: about 25 minutes each, so roughly two hours. Recording the decision: 10 minutes. That is a little over three hours for the range name. The AI itself costs nothing extra if you already pay for a chat assistant; the free tiers of the main assistants are enough for this job. Domain registration is a modest yearly fee that varies by ending and registrar, so look at the price at the checkout before you commit.

If a trademark specialist later reviews the one name you intend to protect, that fee is the real cost of the exercise and is well worth knowing about before you print stock. A range name you have to abandon after 2,000 labels are printed costs far more than any adviser.

Signs the name is working after launch

After launch, look at three things. Do customers spell and say the name correctly when they email or ring? Search the range name and see whether your shop appears on the first page within a few weeks, which tells you the name is distinctive enough to find. And check, once a quarter, the registers again for new marks close to yours, since filings appear over time. If a close match shows up, speak to an adviser before it grows.

Once the names are in, the descriptions and titles follow. Writing Shopify descriptions with AI picks up where this leaves off, and testing an AI title generator shows how to check titles built on your new names before you change live listings. If you are launching several ranges across the year, put the brief, the prompt and the filter sheet in a shared Project so each launch starts from the last one.

For other kinds of shop the example businesses change but the method does not. A pet shop naming its own treat range, a garden centre naming a line of seed packets, or a sports club naming its kit range would all go through the same five stages: brief, thirty names in three styles, written filter rules, four checks in order and a recorded decision. The AI changes how fast you reach the shortlist. It does not change who is responsible for the checks, and it never replaces a qualified adviser when the name is worth protecting. If you want help setting this up inside your own launch routine, that fits within what my AI implementation consultation covers; see the contact page.

Naming questions that come up halfway through

Can an AI product name generator guarantee a name is free to use?

No. A chat assistant does not search trademark registers reliably, even with web search switched on, and can state that a name is unused when it is not. Treat every name it gives as unchecked, and search the registers yourself for each market where you sell before you print labels or buy stock.

Do I need a matching domain for every product name?

Not for every product. A range name or brand name benefits from a matching domain or a clean handle on your store. Individual product names usually live as pages on your existing domain, so check the handle and the search results rather than buying a domain for each.

How many names should I ask the AI for?

Ask for 30 across three styles, then cut hard. Fewer than 20 gives you too little to filter, and over 60 mostly produces near-duplicates. The value is in the filtering rules you apply, not in the volume the tool can produce.

Should I use a made-up word or a descriptive name?

Made-up or unusual words are easier to protect and to find in search, but need more marketing to explain. Descriptive names explain themselves but are often too common to protect. Many small shops pick a distinctive word plus a plain descriptor, such as a coined name followed by 'candle'.

Further reads

• How to Use AI for Etsy Titles, Tags and Descriptions — Titles and tags for a range sold on a marketplace too.
• Shopify Magic vs ChatGPT for Product Descriptions — Compare built-in and general assistants for product copy.
• How to Give AI Your Business Context So Answers Fit — A context block makes naming prompts far more specific.
• Taking a Local Shop Online With AI: Product Listings in a Weekend — List a first range quickly once the names are chosen.
• How to Create Outfit and Styling Posts for a Boutique With AI — Announce the range with styled social posts.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Shopify developer documentation on handles, Shopify help page on URL redirects, WIPO Global Brand Database page and the IP Helpdesk note on TMview (checked 5 Nov 2026).