Shopify Inbox vs Tidio Lyro: Which AI Chat Tool Works Better?

For a small Shopify store, start with Shopify Inbox's AI agent: it is free, works from your catalogue and policies, and takes minutes to switch on. Choose Tidio's Lyro if you need to teach it beyond your store data, take chats from several channels, or want more control. Run the same 20 questions through both before deciding.

That is a starting rule, not a verdict. Inbox wins on cost and set-up time almost by default; whether it wins on answers depends on your products and the questions your customers actually send, and nobody's demo can tell you that. Twenty of your own questions, asked of both bots on the same afternoon, settle it in about an hour.

Follow me on Instagram@sagnikteaches

What each tool is, in one paragraph apiece

Shopify Inbox is Shopify's own chat app for your online store. Its listing describes an AI sales associate that answers buyers and helps them find products. It connects to your catalogue, inventory and policies automatically, lets you set tone and rules, runs in more than 20 languages and can be used with or without the AI agent. The listing shows it as free. Shopify's help centre says Inbox is available on the Basic, Grow, Advanced and Shopify Plus plans, and that the Inbox agent does not work with legacy customer accounts, so a store still on those keeps every conversation with staff. Customers also see a notice that they are chatting with AI, which you cannot remove or reword.

Connect on LinkedInSagnik Bhattacharya

Tidio is a separate chat and automation platform that works with Shopify and other platforms. Lyro is its AI agent, trained on content you give it. Tidio's free plan allows 50 billable conversations a month and, separately, the first 50 Lyro conversations free for life. Paid plans are $24.17 a month (Starter) and from $49.17 (Growth), with Lyro sold on its own from $32.50 a month for 50 conversations. Tidio's page does not say which billing cycle those figures assume, and it shows an annual toggle offering two months free, so check the cycle before comparing.

Subscribe on YouTube@sagnikteaches

The two lead you to different set-up habits. Inbox is a switch-on tool with limited levers. Lyro is a tool you teach, and it rewards time spent on it.

The test: 20 questions from a real shop

Vendor demos use friendly questions. Your customers do not. Build a set of 20 from your own inbox: scroll back through the last three months of emails, chats and Instagram messages and copy the questions as written, typos included. For an independent bookshop, a realistic spread looks like this.

The traps matter most. They are questions where the right answer is "I cannot do that, here is a person who can". A bot that invents a price-match policy has just created a promise you now have to keep.

Run both bots the same way

1. Switch on the Inbox agent at a quiet hour (its assignment setting offers "All conversations", "Only when your staff is unavailable" or "Never") and ask your questions from a private browser window. Set up Lyro on a free Tidio account with the same product feed, policy pages and FAQ text.
2. Ask all 20 questions in the same order and wording to each bot, copying each reply into a sheet.
3. Score every reply 0, 1 or 2: 0 is wrong or invented, 1 is right but unhelpful or incomplete, 2 is right and useful. A hand-over to a human for a question that needs one scores 2.
4. Add two columns: "would I be comfortable if a customer saw this" (yes or no) and "time to set up" in minutes.
5. Run the whole set again a week later, after you have fixed the gaps you found. That second pass shows how much effort each bot needs.

A filled-in scoring sheet

This sheet is a template with invented scores, to show how to read the results. It is not a measurement of either product.

Read the table in three ways. Bot A is better out of the box. Bot B improves a lot once its gaps are fixed, which says it rewards teaching. And the trap row shows Bot B's first pass was risky, so it should not go live until that row is clean. The right choice depends on whether you have an hour a week to tune a bot.

Where each one is likely to stand out

These are predictions from how the tools are built, not results. Test them.

• Product facts and stock: Inbox connects to your catalogue and inventory automatically, so it should have the edge on "is it in stock". Lyro gets product data through your integration and any pages you add, so check how fresh its stock information is.
• Policies: both work from policy text. If your returns policy is vague, both will answer vaguely. Fix the page, not the bot.
• Order status: this is where small-shop bots often fall down, because they need a lookup of order data and a safe way to verify the customer. Shopify's help centre says the Inbox agent looks up order status only after the customer signs in, which is the safe design. Check how Lyro handles "where is my order" in your set-up without asking for details it should not need.
• Recommendations: Inbox describes itself as a sales associate, so it may suggest products from your catalogue more naturally. Lyro will lean on what you have taught it.
• Several channels: Tidio is a multi-channel inbox, whereas Inbox lives in your Shopify store and its email chat. If customers reach you on social channels, that difference decides it.

Cost at three chat volumes

Illustrative maths with the list figures above. Say the bookshop expects bot-handled conversations of 40, 150 and 600 a month.

• 40 a month: Inbox is free (your Shopify plan is already paid). Tidio's Lyro free tier covers the first 50 conversations in total, not per month, then the paid quota starts at $32.50 a month. Either costs nothing for a month or so.
• 150 a month: Inbox is still free. Lyro needs a paid quota. The page's starting price is $32.50 a month for 50 conversations and its calculator runs from 50 up to 1,000 a month, but it does not print a per-conversation rate, so set the calculator to 150 rather than multiplying the starting figure by three.
• 600 a month: Inbox is free; Lyro sits in a mid-sized quota, priced on the same calculator. Compare that with an hour of a person's time: if the average chat takes four minutes to answer by hand, 600 chats is 40 hours.

"Free" Inbox is not free of cost if it gets answers wrong. Put a value on a mistake: one wrong delivery promise that leads to a refund costs more than a month of Lyro. That is why the scoring sheet counts wrong answers as a 0 rather than skipping them.

For the wider bill of AI add-ons on a Shopify shop, the tutorial on how much it costs to add AI to a Shopify store adds up chat, descriptions and automation together.

Teaching Lyro: what you give it

Lyro's quality depends on the material you supply. A first set for a small shop usually has four parts: your policy pages (shipping, returns, privacy), a short FAQ in the customer's words, your product facts in a consistent format and a list of "never answer, hand over" topics. Writing that FAQ is where an AI assistant helps, so long as you check the output.

Here are 30 real customer questions from our inbox and our shipping
and returns policy. Write a customer-facing FAQ answering each
question in two sentences, using only the policy text. If the policy
does not answer a question, write "HAND OVER" instead of an answer.

A realistic fragment of output, illustrative: "Q: Can I change my delivery address after ordering? A: HAND OVER." The assistant marked this as hand-over because the policy said nothing about address changes. That is the useful behaviour: it shows you the gaps in your policy before a customer does. Write the missing line, then re-run.

The same habit helps Inbox. Because it learns from your store content, improving your policy pages and product descriptions helps both bots, and the tutorial on writing Shopify product descriptions with AI is a good place to start.

One Saturday at the bookshop: where a bot earns its keep

Numbers on a sheet are abstract, so picture a Saturday for an independent bookshop with a Shopify store and one person in the back room packing orders. Between 10 am and 4 pm the chat widget receives 14 messages. Five ask whether a title is in stock or when a restock arrives. Three ask about delivery before a date. Two are returns questions. Two ask for a recommendation. One asks for a price match against a large retailer, and one is a complaint about a parcel that has not arrived.

A good bot, in this picture, deals with the five stock questions, the three delivery questions and one of the recommendations on its own: nine of 14, roughly two thirds. The two returns questions get the policy text and an offer of a person. The price-match question and the missing parcel go straight to a human with the customer's words attached, because the bot cannot grant exceptions or look at a carrier's tracking page. The person in the back room replies to those three after packing, rather than breaking off nine times to answer questions about stock.

That is the realistic win: not "the bot replaces support" but "the packer is interrupted three times instead of fourteen". It is also why the trap and hand-over rows in the scoring sheet matter more than the product-fact rows. A bot that answers nine easy questions well and bluffs on the other five is worse than one that answers six and hands over eight.

Writing the rules you give either bot

Both tools let you set rules, and the wording of those rules is part of the test. Write them as short, testable statements rather than a paragraph of personality. A filled-in set for the bookshop, illustrative:

Tone: friendly, brief, no exclamation marks.
Never promise a delivery date. Give the standard range and say
that dates are not guaranteed.
Never offer refunds, discounts or price matching. Hand over.
If a book is out of stock or on back order, say so and offer to
note the customer's email for a restock message.
If the customer sounds upset, hand over at once.
If you are not sure, say "I will pass this to a person" and hand over.

Notice that every line can be tested with a question from the 20. "Never promise a delivery date" is checked by the shipping group; "hand over at once if upset" is checked by a deliberately cross message such as "this is the second time you have lost my parcel". If you cannot think of a test question for a rule, the rule is too vague to be enforced and you should rewrite it.

Inbox lets you set tone and rules in its own settings. In Lyro, the guidance items are the closest equivalent, and Tidio's plan table allows 5 of them on Free, Starter and Growth alike, rising only on the Plus and Premium plans. So choose the five rules that protect you most (no promised dates, no refunds, no price matching, upset customers handed over, honest "I do not know") and fold any extra wording into your FAQ content instead.

The hand-over message that customers actually accept

Whichever bot you choose, what it says when it gives up shapes the customer's mood. "Sorry, I cannot help with that" feels like a wall. A better hand-over states what happens next, by when, and carries the context so the customer does not repeat themselves. A before and after, illustrative:

Before: "I am unable to assist with this request. Please contact support."

After: "I cannot change an address once an order is packed, so I am passing this to a person who can check. They will reply by email before 5 pm today, and I have included your order number and new address so you will not need to repeat it."

Test the hand-over in your 20 questions as well: after each hand-over, ask whether the human receiving it would have enough to reply without asking the customer anything. If they would not, adjust the bot's guidance to include order number, the question and the customer's contact detail every time. Inbox handles part of this for you: its help centre says the agent's context carries over when staff take the chat, and when no staff are available it gives the customer your store's sender email address instead, so check that address is one somebody actually reads. And decide in advance who is on duty to receive hand-overs and what the promised reply time is, because a bot that promises a reply you cannot deliver is worse than no bot.

Three failures to watch for in the first fortnight

The confident wrong answer. Say a customer asks whether a book arrives before a birthday, and the bot says "yes, two days" because that is the standard shipping time, ignoring that the item is on back order. Guard against it by adding the back-order rule to your guidance and by testing a back-ordered product on day one.

The polite dead end. The bot keeps apologising and rephrasing without ever offering a person. Count the turns before a hand-over in your test; more than three is too many for a frustrated customer.

The hidden meter. Tidio counts a Lyro conversation the first time Lyro replies in a new conversation, so every test chat you run on the live widget uses allowance, and the free 50 can vanish during set-up. Watch the usage page for the first week so the first bill is not a surprise. The counting is fair in one respect: Tidio's FAQ says a conversation counts once however many times Lyro replies, even if the customer leaves the chat window and comes back to it later.

A decision rule you can apply in a morning

1. Run the 20 questions through Inbox first. If it scores 34 or more out of 40 and has no trap failures, stay with it and stop.
2. If it scores below that, or you need chats from channels outside your store, set up Lyro on the free tier, fix the gaps and re-run.
3. If Lyro reaches 34 or more after one round of fixes, price the block you need. Choose it only if the extra score is worth the monthly fee and the weekly maintenance.
4. If neither reaches the bar, keep a human on chat and use the bot only to draft replies for you to approve. A draft-only mode loses no customers.

For a broader field than these two, the best AI chatbots for small Shopify stores compares more options, and the comparison of Tidio against Intercom explains what a step up from Tidio looks like if the store outgrows it.

What to measure after it goes live

Pick three numbers and read them weekly for a month: the share of chats the bot closed without a human, the share of closed chats where the customer then emailed anyway (a sign the answer failed), and the count of complaints that mention the bot. If the second number is above about one in five, the bot is deflecting rather than helping. Re-run the 20 questions monthly and add every new question that stumped it.

A quick example. A bookshop sees 200 chats a month, the bot closes 120, and 30 of those customers email again. That is 90 genuinely resolved, 45 percent of all chats, and a failure rate of 25 percent on closed chats. Fixing the five most common repeat topics in the knowledge base should cut that next month. Keep the log; it is also the best evidence of whether the fee is worth paying.

Questions before you switch a chat bot on

Is Shopify Inbox's AI agent really free?

Shopify's app listing shows Inbox as free, and Shopify's help centre says Inbox is available on the Basic, Grow, Advanced and Shopify Plus plans. You still pay for your Shopify plan, and the Inbox agent does not work with legacy customer accounts, so a store on those keeps chats with staff until it moves to new customer accounts.

Does Tidio's free plan include Lyro?

Yes, but only the first 50 Lyro conversations, and the page describes that allowance as lifetime rather than monthly. After that you need a paid Lyro or plan tier. Free also limits you to 1 Lyro action and 5 guidance items, so treat it as a trial, not a permanent setup.

Can either bot take actions like refunds or order edits?

Neither should be trusted with refunds in a small shop without a human approval step. Tidio's free tier includes one Lyro action, with full access only on a paid Lyro quota, and Shopify's help centre says the Inbox agent can recommend products, add items to the cart and look up order status for signed-in customers. Start with answers only, then add actions slowly and test each one.

What should the bot do when it does not know?

It should say so and hand the conversation to a person with the question attached, not guess. In your test, count every confident wrong answer as a serious failure and every honest hand-over as a pass. A bot that is right 80 percent of the time and admits the rest is safer than one that is right 90 percent and bluffs.

Further reads

• AI Email Triage for Shared Inboxes: Sales, Support, and Invoices — Sort the emails that chat cannot answer.
• AI Ticket Triage: Tag, Route, and Prioritise Support Requests — Tag and route what the bot hands over.
• How to Manage Your Inbox With AI: Triage, Drafts, and Follow-Ups — Triage the emails the bot hands back to you.
• Stocky Alternative for Shopify: Purchase Orders and AI Reordering — Another Shopify change affecting small stores this year.
• How to Sync Stock Across Shopify, Amazon and eBay Automatically — Accurate stock is what lets a bot answer availability questions.
• Startup AI Tools for Growth: 8 That Earn Their Subscription — Eight growth-focused AI tools judged by the job they do, ranked by payback, with a 14-day trial rule and a cancel list.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Shopify Inbox app listing, Shopify help centre (Inbox agent article), Tidio pricing page and FAQ, read 4 and 5 November 2026.