AI Pre-Purchase Chatbots: Turning Product Questions into Orders

Pre-purchase and post-sale chatbots are different products

A lot of stores treat “chatbot” as one thing and install whatever’s cheapest. Pre-purchase and post-sale are different jobs with different goals, and mixing them up wastes money.

A post-sale support bot (the kind we covered in our multilingual customer service guide) answers questions from people who already ordered: where’s my package, how do I return this, how long is the warranty. Its main metric is resolution rate and response time. It’s a cost-saving tool.

A pre-purchase bot answers questions from people who haven’t ordered yet: does this run small, will this material irritate my skin, can it arrive by Friday. Its main metric is conversion rate and average order value. It’s a revenue tool, and it lives in a different place too: product and collection pages instead of order status pages, right where a shopper decides whether to click add to cart.

Running both off the same knowledge base backfires fast. I’ve seen stores plug a support bot’s FAQ into the product page, and when a shopper asks “will this jacket shrink in the wash,” it responds with “please provide your order number to check shipping status.” That’s a lost sale over a mismatched tool. Decide which stage you’re solving for before you pick anything.

How much conversion lift is actually real

Industry benchmarks put AI chatbot conversion lift somewhere between 15% and 35%. Rep AI’s own case studies cite a 10-30% conversion lift within the first 30 days after install. That’s a wide range because it depends heavily on traffic quality, price point, and how complex your catalog is.

One data point worth flagging: shoppers who engage with an AI chat get to purchase 12.3% of the time, versus 3.1% for shoppers who don’t engage, roughly 4x. That doesn’t mean installing a bot quadruples your store’s conversion rate. It means the small slice of visitors who actually open the chat window convert far better, and most visitors never open it at all.

Proactive triggers (the bot pops up after 20 seconds of hesitation) outperform passive ones (waiting for the shopper to click) by another 20-25%. That’s why Rep AI and Tidio Lyro both push “proactive” messaging as a headline feature, since a bot that waits to be clicked leaves most of its potential conversion on the table.

AOV gets a lift too. Stores using bot-assisted product selection see AOV climb 10-15% on average, and stacking in upsell and cross-sell recommendations pushes that to 8-20%. Add both together on a mid-traffic store and the extra revenue can be several times the subscription cost. That math only works if your catalog is complex enough to need guidance. A store selling one t-shirt in five colors won’t see much from any of this.

Comparing Tidio Lyro, Rep AI, Gobot, and Octane AI

Tidio Lyro’s biggest trap is the pricing structure. Free and Starter plans include only 50 Lyro AI conversations total, and that’s not per month, it’s total, ever. Once they’re gone you’re paying add-on fees, and the real bill often ends up double what the homepage number suggests. It’s a solid pick if you’re already running Tidio for basic support and want to layer AI Q&A on top, but the free tier won’t carry real pre-purchase conversion on its own.

Rep AI is the priciest and most aggressive of the four. Its pitch is proactive intervention: it tracks scrolling, dwell time, and hesitation on the product page, then jumps in with a targeted suggestion. One merchant case study cites over $69,000 in incremental sales in a single month and a conversion lift near 25%, but that kind of number tends to show up at stores that already had meaningful traffic and AOV going in. If you’re only pulling a few thousand visits a month, a $250/mo starting price is hard to justify before you’ve proven the model works for you.

Gobot’s free tier is its real advantage. Run the free 5,000-engagement allowance first, see what the data actually shows for your store, and only add budget once you have proof. Its quiz mode is simple enough that a small team can build one in an afternoon, which makes it a low-risk starting point.

Octane AI takes a completely different approach: quizzes that ask directly about skin type, budget, or use case, then match products against the answers. Beauty and supplement brands lean on this “consult first, recommend second” flow because customers expect to answer a few questions before getting product suggestions anyway. The credit-based pricing is hard to estimate upfront, so run a month on the Starter tier before deciding whether to scale up.

How the product recommendation logic actually works

The four tools split into two recommendation approaches. One is signal-based, like Rep AI: the system watches what a shopper does on the page: which product image they linger on, when their scroll speed suddenly slows, when the cursor hovers over “add to cart” and pulls back. Those signals get read as hesitation, and the bot proactively offers something like “want help comparing these two?”

The other is preference-based, like the quiz flows in Octane AI and Gobot: ask a few direct questions: budget range, use case, skin type or body type. Then match the answers against the product catalog. This tends to convert more reliably because self-reported preferences beat inferred behavior, but it costs setup time to write good questions, and shoppers bail if the quiz runs too long.

The stores getting the best results combine both. Collect baseline preferences through a quiz, build a lightweight customer profile from it, then let the chat bot reference that profile during later browsing. If a shopper flagged sensitive skin in the quiz and later views a product with alcohol in the formula, the bot can flag it directly: “this one contains alcohol. Based on what you told us, here are a few alternatives.” Right now only some tools connect the two layers cleanly. Tidio and Octane AI integrate reasonably well together; Rep AI and Gobot mostly operate in their own silos.

Recovering abandoned carts through chat

Industry data suggests well-configured exit-intent chat recovers 10-15% of abandoned carts, with the most carefully tuned deployments claiming as high as 35%. Numbers at the high end almost always come paired with a discount or free-shipping incentive. Chat alone rarely gets you there.

Timing matters more than most teams assume. The highest-converting moment is when the cursor moves toward the browser’s close button, or when a shopper sits idle on checkout for over a minute. Trigger too early and you interrupt normal browsing before the shopper has even decided to leave, which tanks conversion instead of saving it.

Layer the message instead of leading with a discount. The first prompt should ask what went wrong, something like “run into an issue at checkout?”, because a lot of abandonment comes down to unexpected shipping costs or a missing payment method, and fixing that specific problem converts better than a coupon and doesn’t cost you margin. Save the discount for a last resort if the real issue can’t be resolved.

Compared to email and SMS recovery, chat’s advantage is immediacy. The shopper is still on the page and the objection gets handled on the spot instead of waiting for tomorrow’s email. Its weakness is reach: most shoppers close the tab before any popup fires. The best setups run all three together, with chat catching the moment and email/SMS handling the one-to-three-day follow-up window.

Setting this up on Shopify without breaking anything

Check that your pre-purchase and post-sale bots don’t overlap in where they trigger. Product, collection, and cart pages get the pre-purchase bot; order status and account pages get the post-sale bot. Let both fire on the same page and shoppers get annoyed enough to close the whole chat widget.

Get your product data in shape before launch. All four tools read your Shopify catalog (titles, descriptions, variants, stock) to answer questions and make recommendations. If your product copy says “one size fits most” with nothing more specific, the AI can’t give a real sizing answer no matter how good the model is. Fix the descriptions on your core SKUs first instead of hoping the AI will fill the gaps.

Pilot on a handful of pages before rolling out site-wide. Pick your three to five highest-traffic product pages, run the bot for two weeks, and look at both conversion data and customer feedback before deciding to expand. Gobot’s free tier is built for exactly this kind of test.

Watch for overage billing, especially on Rep AI’s $12-per-1,000-visits charge above your plan limit. A traffic spike during a sale can push your bill well past what you budgeted, so check your plan tier before any big promotion instead of finding out after the invoice lands.