Mike HeapMike is an experienced Product Manager who focuses on all the “non-development” areas of My AskAI, from finance and customer success to product design, copywriting, testing and more.
Table of Contents
- Zipchat at a glance
- What is Zipchat?
- How easy is it to set up Zipchat?
- What channels does Zipchat work in?
- What are the limitations of Zipchat?
- What knowledge sources can I train Zipchat on?
- What features does Zipchat have?
- How do I improve Zipchat's responses?
- What resolution rate can I expect from Zipchat?
- What AI model does Zipchat use?
- What languages does Zipchat work in?
- How secure is Zipchat?
- Who is using Zipchat?
- How much does Zipchat cost?
- The tiers
- What the meter actually counts
- The App Store and the website sell different price lists
- Contract terms
- A worked example
- Does Zipchat have a free trial?
- Is Zipchat worth it?
- How does Zipchat compare with My AskAI?
- What are the pros and cons of Zipchat?
- Pros
- Cons
- FAQs
Created time
Aug 6, 2026 11:54 PM
Title length (<60)
Author
Last optimised
Ecomm?
Image
Publish date
Jul 30, 2026
Video
Slug
zipchat-complete-guide-2026
Featured
Type
Article
Ready to Publish
Ready to Publish
Zipchat starts at $49 a month on its own site, and $0 if you install it through Shopify. It meters AI replies rather than tickets, and that one choice decides most of the cost math.
Someone has asked you to put AI on the store chat, and Zipchat keeps coming up.
An hour on their site later, the picture is still fuzzy. The homepage sells revenue, the docs sell support, and five of their own pages give the automation rate three different names and four different numbers.
That's a hard place to buy from. You have to say yes or no this week, and the vendor's own numbers won't sit still long enough to compare.
I'm Mike, co-founder of My AskAI. We help 200+ ecommerce and SaaS businesses run AI customer service, so I obviously can't be completely impartial here. By the end you'll know what Zipchat is built to optimize, what it costs at your volume, and whether that's your problem.
Zipchat at a glance
Fact | Detail |
What it is | Standalone, sales-first AI chatbot for ecommerce |
Headline price | $49/mo Starter, up to $499 Scale, $999+ Unlimited |
Free option | Permanent free plan: 120 AI replies/mo, Shopify only |
G2 rating | 4.8/5, from a small set of 9 reviews |
Best for | DTC stores losing sales to unanswered pre-purchase questions |
Not a fit for | Support-led teams, or non-Zendesk helpdesks |
Biggest limitation | Three labels, four figures, one handling rate |
Our verdict | 7/10 |
My AskAI is an AI support agent that runs inside the helpdesk you already pay for: Zendesk, Intercom, Freshdesk, Gorgias and HubSpot. We bill a flat rate of around $0.10 a ticket, and on Shopify our pre-built Task recommends products and checks live stock.
What is Zipchat?
TL;DR: Zipchat is an AI chat agent for ecommerce stores that is built around selling, with support answering layered on top. It is not a helpdesk, and it doesn't live inside one.
Zipchat is an AI agent that lives on your storefront, installed as a chat bubble through a Shopify app embed or a single script anywhere else. It brings its own Conversations inbox for your team to work in, and it connects to one helpdesk, Zendesk, so on any other ticketing platform that's a second place to answer from. If you're mapping this against your stack, that's the first thing to write down.
Its own docs put the job description like this:
"an AI Agent for eCommerce — a next-generation system that helps Shopify and online stores sell more, support faster, and communicate smarter"
The same page lists three core missions in this order: Sell More, then Support Smarter, then Learn and Improve. Support is explicitly second, and I don't think that ordering is an accident. The homepage doesn't hide it either, running "Turn your support into a profit center" over a hero stat of "97% resolved without looping your team".
All the social proof on that page is money. A band of stat cards credits Tropicfeel with over 100k€ from the AI chatbot, Burger Motorsport with over $1m, and Shelly with 10x return on investment. Nowhere on that page is there a CSAT figure or a first-response time.
The Shopify listing launched in May 2023 and carries the Built for Shopify badge (Shopify's own vetting bar, which counts for something on an app you're dropping into your theme). Zipchat's about page claims 2,000+ companies across 40+ countries, both self-declared.
G2: Zipchat scores 4.8/5 from 9 reviews on G2. "Unlike traditional chatbots that rely on rigid, pre-programmed flows, ZipChat truly understands our entire online catalog. It answers complex customer queries about product specifications, availability, and recommendations flawlessly, mimicking a highly trained in-store sales representative." via a G2 reviewer (consumer goods, small business).
Nine reviews is a small set, so I'd read the quote as one merchant's experience and the 4.8 as a weak signal.
My AskAI runs inside the helpdesk your team already works in, so tickets, macros and routing stay where they are. The AI answers on the same threads your agents do.
How easy is it to set up Zipchat?
TL;DR: Install from the Shopify App Store and it trains itself off your catalog and site content in minutes. The work sits in the guardrails afterwards.
The documented Shopify flow runs to six steps. Install the app, create an account, accept Shopify's terms, pick a plan to start the trial, then train the agent in the dashboard.
The bubble goes live through a theme app embed, or a JavaScript snippet in your head tag on any other platform. I'd have your theme developer nearby for that second route if you've customized heavily.
Training starts on its own, with the agent crawling your site for product details and policies.
Merchants back the speed up, and one G2 reviewer puts it bluntly:
"I've worked with competitors before, and I can say Zipchat AI is the easiest setup I've ever done. It's a 10 out of 10."
The pricing page gives two setup times in two places: "Setup in 2 minutes" on the CTA, 5 to 10 minutes in the FAQ. Every new plan also includes a free onboarding call, which is unusually generous at $49 a month. Take it.
The install is the fast part. The onboarding page still offers a video marked "(COMING SOON)". Sites behind Cloudflare can block the crawler and need a firewall rule, which Zipchat keeps a dedicated article for.
And a B2B reviewer on G2 flagged that the ecommerce assumptions make setup harder when you aren't a store.
My AskAI installs from your helpdesk's own marketplace in under 10 minutes, no developer needed. The Shopify integration covers product, order and customer lookups out of the box.
What channels does Zipchat work in?
TL;DR: Five channels on every paid tier: website chat, WhatsApp, Instagram DMs, Facebook Messenger and email. No voice, no SMS, and exactly one helpdesk integration.
Nothing here is tier-gated, which is unusual at this price. The chat widget comes in three flavors (a bubble, a standalone link-to-chat page and an embedded iframe), all covered in the channel docs.
WhatsApp does double duty: inbound support through the Business API, plus outbound marketing campaigns. Those sends carry Meta's own per-conversation fee on top of your plan, which Zipchat prices at around $20 per 1,000 people reached.
There is no voice or phone channel and no first-party SMS, so I'd check that against your own inbound mix. The Shopify listing's checklist does tick a "Phone" box, but that's a tick-box on a marketplace form with no documentation behind it. Store platform coverage is broader: Shopify first and deepest, then WooCommerce, BigCommerce, Magento, PrestaShop and Wix.
Helpdesk coverage is the thin spot. The integrations page lists exactly one helpdesk, Zendesk.
Gorgias, Intercom, Freshdesk, HubSpot and Klaviyo appear on Zipchat's site only as alternatives pages, which is to say as competitors to replace. Four of those five are helpdesks we run inside.
Zendesk itself comes with a caveat. The integration page says email answering "Works on any Zendesk plan" while live chat "Requires Zendesk Suite Professional or above". Its own FAQ then says the integration connects at account level and needs no specific tier.
Both sentences are live right now, so I'd get the answer in writing. The Shopify listing muddies it further, naming "Gorgias, Zendesk, Re:amaze" in its Works with block while the integrations page names neither Gorgias nor Re:amaze.
My AskAI answers whatever channels your helpdesk routes to it, including social. It hands over to a human inside that same helpdesk, with a summary attached so your agent picks the conversation up cleanly.
What are the limitations of Zipchat?
TL;DR: The biggest is that the support-handling rate appears on five of Zipchat's own pages under three labels and four figures, and no page defines any of them. Second is the reply meter, which charges support chatter at the same rate as a sale.
Most of what follows is baked into how Zipchat is built. I'd read all four as design choices.
Start with the number that won't hold still. Five Zipchat pages publish a handling stat, all live on 6 August 2026, and between them they use three different words for it:
Page | The claim |
"97% resolved without looping your team" | |
"97%+ Support cases resolved automatically" | |
"Zipchat handles 90%+ of inquiries automatically, with some merchants getting close to 99%" | |
"the AI resolves 70% to 85% of tickets automatically" | |
95%+ deflection, on Zipchat's separate SaaS product line |
That's five pages and four figures, because the two 97s are one claim printed twice. The 95%+ belongs to Zipchat's code assistant for SaaS teams, so on the four ecommerce pages you're choosing between 97%, 90%+ and 70% to 85%.
I'd worry about the words more than the count. Resolved, handled and deflected are three different events, and no page says which one it means.
The customer service page puts two of those words on one range. Its prose says the AI resolves 70% to 85%, and the comparison row carrying the same figures is headed Deflection rate.
Every figure in that table carries the same attribution, "Zipchat analysis of active client deployments", with no methodology, no sample size and no date range. I'd ask for the sample size on a call.
The automate-support page also states that fewer than 3% of conversations escalate. Then it works the example as "At 97%+ deflection, 1,000 monthly contacts become about 90 for your team".
Three percent of a thousand is thirty. Ninety implies about 91%, which is the number on the pricing page.
The second is the meter. It counts replies rather than conversations, and a reply is any message the AI sends on any channel.
Zipchat's own tier math assumes roughly 2.4 replies per conversation (our own chat billing counts every two AI replies as one credit, so the math will feel familiar). So a support back-and-forth costs the same as a chat that ends in a sale. Measure your own replies-per-conversation before you pick a tier.
Third is the helpdesk count, which is one. I'd check that before anything else here, because if your tickets live in Gorgias, Intercom, Freshdesk, HubSpot, Help Scout or Kustomer, Zipchat answers on your storefront and your ticketing carries on separately.
Fourth, the knowledge layer only reaches your own content. There are no wiki or cloud-drive connectors, no way to link a live document and no documented training on your past tickets. We hear about that last one on most onboarding calls we run.
A verified G2 reviewer describes the workaround:
"Also, the bot can't connect to live documents like Google Sheets, so I have to upload PDFs daily to keep product info updated."
My AskAI's Knowledge reads your help center, your website, uploaded files and cloud sources like Google Drive, Notion, Confluence and SharePoint. Self-Learning then drafts new articles by comparing the AI's reply to the human's.
What knowledge sources can I train Zipchat on?
TL;DR: Seven documented sources, all of them URLs, files or pasted text, plus your Shopify catalog automatically. Caps run from 1,000 training pages on Starter to 300,000 on Scale.
The knowledge base docs list seven ways in, and they're all variations on the same idea:
- Automatic Crawling, which "scans all your site pages and adds them to the Knowledge Base"
- Single Page, for one URL, subdomain or external site
- Multiple Pages, which scans everything linked from a main URL
- PDF Uploads, for guides, charts and manuals
- Blank Text, for content typed straight in
- CSV with Links, for a list of URLs
- Sitemap (XML), for a full site-structure crawl
On Shopify the catalog comes in on its own, so the agent can answer on products, pricing and availability from day one. There's an eighth source too, off that list: every chat correction you write is appended as a question-and-answer pair.
The caps scale with the tier and they're generous: 100 training pages on the Shopify free plan, 1,000 on Starter, 15,000 on Growth, 75,000 on Pro and 300,000 on Scale. One published customer, moto24.de, reports 30,000 pages in its knowledge base, so count your own pages before you choose a tier.
No refresh cadence for the website crawl is documented anywhere, and a G2 reviewer has noticed:
"Zipchat AI should update and reindex the pages periodically to avoid providing obsolete information."
I'd put a manual recrawl in someone's calendar, because nothing in the docs says Zipchat will do it for you.
If your knowledge isn't written down yet, we build starter knowledge from your last 5,000 resolved conversations with Train on Historic Tickets, more on request. A team with no help center can still get an agent live.
What features does Zipchat have?
TL;DR: Three layers: an autonomous agent that sells and answers, a takeover layer for your team, and an action layer you can extend without code. Every AI feature is on every paid tier.
I'd start with the pricing table. Agentic capabilities, agentic search, the Ask AI bar, follow-ups, human escalation, proactive chat, WhatsApp marketing, analytics and API access all show a tick on Starter through Scale. Team seats are unlimited on every tier too, so the plans gate capacity while the features stay flat.
The pre-built action set is deliberately small: knowledge search, escalate conversation, order lookup, create discounts on Shopify, and skip conversation. Anything beyond that is a skill you build, no-code, in what Zipchat says is under an hour. I'd budget longer for the first one.
You describe the action in plain language, "Paste your platform's API, webhook, or MCP", then set the guardrails. MCP support is rare in this category, so if you already run a server it's the quickest route in.
Behavior runs off a four-layer prompt stack, which I'd get written down before you go live: a core prompt for personality and rules, contact info for escalation, additional instructions for sales logic, and per-tool instructions. Tools switch on or off per channel, so switch off anything you don't want it doing on day one.
The sales side is where the product spends its energy. Proactive chat fires context-aware messages off live browsing behavior, campaigns broadcast to segmented lists, and the agent can mint single-use Shopify discount codes mid-conversation.
The support side covers order tracking, delivery estimates, returns, refund guidance and escalation into Zendesk. Your team takes over mid-conversation from the inbox. I'd rehearse that handover in the sandbox before you go live.
Shipping speed is fast and checkable, which I always take as a good sign. The changelog carries fifteen entries in the six weeks to 6 August 2026, including agent memory that builds persistent customer profiles.
My AskAI runs actions through Tasks and Tools. We build Tasks that describe what to do in plain language, and Tools that connect the systems. On Shopify a pre-built Task acts as a sales agent, recommending products and reading live inventory and variant stock, and cart, checkout-link or discount actions are yours to wire with Tasks and AI Actions. Each action runs fully autonomously or proposes for approval, your call.
How do I improve Zipchat's responses?
TL;DR: Write corrections for what it knows, edit the prompt stack for what it does, then watch a thumbs-rating loop. It's reactive and per-incident: you fix answers after customers hit them.
Corrections are the main lever, and one person on your team should own them.
When the agent answers badly you write the right answer, and Zipchat appends it as a question-and-answer pair. One documented boundary: corrections only load on an information search, so they don't reach order tracking or escalation.
Corrections change what the agent knows, and its behavior comes from the prompt stack. I'd keep that split in your head, because mixing them up is where tuning goes wrong. The corrections docs also document where a correction fails.
Conflicting information elsewhere on your site can override a correction, and the troubleshooting list ends on checking whether your own pages have gone stale (we run into the same thing constantly).
Then you read the numbers. Have this screen open through your trial. The dashboard documents ten tiles: sales including a conversation, total orders, chat-to-sale conversion rate, average order value, chats started, total replies, great replies, good replies, bad replies, and the pages chats start from.
There is no CSAT tile, no first-response time, no resolution rate and no deflection rate. If support is what you're measured on, I'd work out what you'd report off this dashboard before you buy.
The reply ratings are the nearest thing to a quality score. Zipchat's definition of a Good Reply:
"Replies that didn't receive ratings—neutral but successful interactions."
An unrated reply counts as a success.
I'd ask about four more gaps. There's no report clustering the questions the agent couldn't answer, no bulk replay against your historical conversations, no confidence tuning and no staged rollout. On a large catalog, working around all four is an ongoing job for someone.
What resolution rate can I expect from Zipchat?
TL;DR: Zipchat publishes four figures under three labels and defines none of them. Its own named customers cluster lower, around 75% to 96%, and no independent source has verified any of it.
Zipchat publishes nothing on what "resolved" means in its system, and the product exposes no resolution metric at all.
Of its four published figures, the one I'd quote is the 70% to 85% range on the customer service page. It's the lowest of the four and the only one given as a range. I'd plan against that and treat anything above it as upside.
Across the 32 published success stories, Almanac Planting reports 96%+ of support automated, CFS a 75% reduction in workload, and Conte Scarpe Moda under 20% escalated. That band runs 75% to 96%, under the 97% on the homepage.
The stories don't always agree either. Family Nation appears at 80% automated on the capabilities page and at 97% of chats on the pricing page, both live today. I'd ask a salesperson which one is current.
Nobody outside Zipchat has checked any of it. One independent reviewer says as much: "I do not have a measured conversion lift from either site, so I am not claiming one."
For where the field sits, our aggregate of 195 rated deployments across 38 vendors puts the median AI handling rate at 70%. The range runs 15% to 98%. Read it as an aggregate and don't rank vendors on it: every vendor defines its own metric, and published figures are self-selected wins.
That explains Zipchat's spread. In our data the label moves the number more than the product does: "resolution" sits at a 72.5% median, "deflection" at 70% and "automation" at 61%.
Conversion, deflection and resolution measure three different things. That's how 97% resolved and 70% to 85% deflection sit on one website.
I've written about this before: any vendor can hit 90% resolution if reaching a human is hard enough.
My AskAI sits at 72%+ on a rolling 30-day basis across our customer base. We count a conversation resolved when the AI handled it without escalating to a human, and we make asking for a person deliberately easy. Insights then scores every conversation for AI CSAT.
What AI model does Zipchat use?
TL;DR: Unknown, by policy. Zipchat states in writing that it doesn't name its models publicly, and its subprocessor list names cloud providers but no AI provider.
Zipchat is vague about its models on purpose, which I prefer to a hint. Its AI platform page says "Zipchat is built on leading frontier AI models", then adds "We do not name specific models publicly."
The terms confirm third parties are involved without naming them, stating that the services use third-party AI providers operating under zero-data-retention policies. The privacy policy names AWS and Google Cloud for infrastructure, plus the usual analytics and billing tools, and no LLM provider at all. The full list is available on request, so ask for it if your reviewer needs named providers.
Third-party directory profiles put a GPT-4 label on it, and the Shopify listing names ChatGPT and Claude. Neither is a Zipchat statement about what serves a customer reply, so treat the model as unknown.
You can't pick or pin a model, and no versioning policy is published, which I'd flag if you run regulated content. They publish one performance number: standard queries come back in under three seconds.
The privacy policy is specific about training data:
"We do not use End-Customer Conversation Data to train shared or foundation AI models, and the AI providers that process this data on our behalf do not use it to train their models."
The terms are broader, granting Zipchat a license to use customer data "to improve Zipchat's own proprietary models". Those two reconcile only if you read both legal pages.
My AskAI names its models: a suite led by OpenAI, with Google and Anthropic on the jobs where they test better. We publish the list and A/B test the mix.
What languages does Zipchat work in?
TL;DR: 95+ languages, auto-detected from the first message, on every tier with no upcharge and no configuration.
Languages are where I'd rate Zipchat highest. The pricing FAQ says "Zipchat detects your customer's language and replies in it from the first message. Any language, no setup on your end".
The docs put a number on it at "95+ languages", and the plan table shows all languages on Starter through Scale (no upcharge anywhere in the ladder).
Detection runs per message, so nothing hangs on where the customer is writing from.
Your knowledge base is built from your crawled site, so an English-only help center means the agent answers in Italian or German off English source material. Zipchat publishes nothing on how that step behaves or whether quality drops.
There's also a fresh reliability note in the changelog. On 3 August 2026 they shipped an entry headed "AI: Improved language recognition", described as making recognition "more stable, reducing cases where the AI would respond in the wrong language".
I'd test your top three languages during the trial anyway.
My AskAI covers 95 languages, auto-detected per message, replying in the customer's language by default.
How secure is Zipchat?
TL;DR: GDPR and EU AI Act compliance are claimed on the badge, the privacy policy is substantive, and no SOC 2 or ISO 27001 certification is published anywhere.
Zipchat publishes no SOC 2 report, no ISO 27001 certificate, no trust portal and no penetration-testing statement, on the marketing site, in the docs or on either legal page. For an SMB product at $49 a month I'd call that fair enough. It's a hard stop if procurement hands you a checklist.
It claims regulatory compliance, and the timing deserves credit. A "GDPR & AI Act compliant" badge sits on the homepage and pricing page, and the EU AI Act became fully applicable on 2 August 2026. Few vendors in this category flag it at all.
The badge is self-declared, and the pricing FAQ's "bank-level encryption" is marketing language, with no TLS version or key length behind it.
The privacy policy is the best security document they have, and I'd warn your reviewer that it is also the ceiling of what Zipchat publishes. Encryption is stated in transit and at rest, and the framework list covers GDPR, CCPA, LGPD and PIPEDA. A DPA comes on request, so I'd ask for it in week one.
ZIPCHAT INC is a US company in Wilmington, Delaware, leaning on standard contractual clauses for EU transfers. No EU data-residency option is published.
Chat logs are kept for the period the merchant configures, then anonymized or deleted. You set that window yourself, which I don't see often at this price.
Access control is thinner. There are roles and unlimited seats, but no SSO, SAML, SCIM, MFA or audit log is documented anywhere. Treat it as undocumented, and ask.
The terms round it out. I'd read them before signing. There's no SLA, only commercially reasonable efforts toward 24/7 availability, and liability is capped at one month of fees.
The AI-output disclaimer is broad too, making the customer "solely responsible for reviewing, verifying, and approving all AI Outputs before use". That obligation sits with the customer even at the 97% resolution rate the homepage advertises.
My AskAI is SOC 2 Type II certified and GDPR compliant, with AES-256 encryption at rest, TLS in transit and isolated containers per customer. The trust center is live and you can read it today.
Who is using Zipchat?
TL;DR: Mid-sized DTC brands, heavily European and particularly Italian, with 32 published success stories and no enterprise logo among them.
The homepage logo bar names around eighteen brands, and I recognized more of them than I expected to. Jackery in portable power and Police in eyewear are the largest, with Shelly and Tropicfeel the most recognizable DTC names. Caleffi, Eprice, Supermedia, CFS and B-exit are all Italian, which is where the company's commercial network sits.
The success stories library is unusually large for a company this size: 32 studies across fashion, beauty, pets, electronics, supplements, food, sport and automotive. Ask for a reference in your own sub-vertical (they have enough studies to have one). A few of the headline claims:
Customer | Sub-vertical | Zipchat's claimed headline |
Burger Motorsports | Automotive | $1m+ from the AI chatbot |
Tropicfeel | Travel gear | 16% chat-to-sale conversion |
The Pigment | Beauty | 33% chat conversion rate |
None of those headline claims is a controlled test. The dashboard metric behind them is "sales including a conversation", which counts orders from customers who chatted, and no holdout or A/B methodology is published.
An independent reviewer makes the same point:
"Whether the lift is causal or correlational (high-intent visitors gravitate toward chat anyway) is a fair question to test on your own catalog before treating the number as a guarantee."
A mid-market G2 reviewer in automotive reports the outcome you'd plan around:
"Zipchat has reduced the frequency and volume of contact to our Customer Service and Technical helpline."
No enterprise logo carries a study, and no named customer sits above roughly mid-market. Some of the growth came from deals, too: a sold-out AppSumo lifetime offer and a referral program paying $50 to $2,000. I'd keep that in mind when you read the 2,000+ company count.
Three of ours, with the numbers published. Edel Optics runs 79% AI resolution and 92% AI CSAT across 4,067 tickets on Zendesk. YouGarden saves 965 hours a month across 11,785 tickets on Freshdesk, and Swytch resolves 4,050+ tickets a month with AI at 81%.
How much does Zipchat cost?
TL;DR: $49 to $499 a month, with Unlimited from $999. The meter counts AI replies, so a conversation is several, and going over costs roughly twice the included rate.
The tiers
Every price is from Zipchat's pricing page, current on 6 August 2026. The conversation column is my own math on Zipchat's assumption of about 2.4 replies per conversation.
Plan | Price/mo | AI replies/mo | Approx. conversations | Training pages | Stores |
Free (Shopify App Store only) | $0 | 120 | ~50 | 100 | — |
Starter | $49 | 500 | ~200 | 1,000 | 2 |
Growth | $129 | 1,500 | ~600 | 15,000 | 5 |
Pro | $249 | 3,000 | ~1,250 | 75,000 | 10 |
Scale | $499 | 6,000 | ~2,500 | 300,000 | 20 |
Unlimited | from $999 | Unlimited | — | — | — |
Unlimited is the only tier behind a demo, an annual toggle advertises 20% off, and nothing is charged per seat or per resolution. That last part I like.
What the meter actually counts
Zipchat defines a reply as "any response the AI sends to a visitor or customer, whether in the chat widget, email, or a connected messaging channel like WhatsApp or Instagram". Each outbound message is one, so a single conversation is several.
Overage is the sharpest edge. Zipchat charges "Extra +$49 for additional 250 replies", which is $0.196 a reply against an effective $0.098 on Starter and $0.083 on Pro and Scale.
An independent reviewer ran the same math: "At full included usage, Starter works out to 9.8 cents per reply. The first overage block costs 19.6 cents per reply, exactly twice that rate."
Zipchat's own FAQ concedes it, saying upgrading is almost always more cost-effective. The move is always up a tier, so I'd size the plan off your busiest month.
The App Store and the website sell different price lists
Zipchat's own site | Shopify App Store | |
Free plan | Not offered | 120 AI replies/mo, 100 training pages |
Starter, Growth, Pro | $49 / $129 / $249 | $49 / $129 / $249 |
Scale | $499 | Not listed |
Unlimited | from $999 | Not listed |
Zipchat's own site never shows the free plan, and the App Store listing never lists Scale. The FAQ says features are identical either way. That holds for the tiers both places show; which tiers you see depends on where you look.
Contract terms
Month to month, no contracts, no cancellation fees, which is what you want at this price. The 30-day money-back badge sits under every tier, including the annual toggle. But terms section 3.3 restricts it to monthly subscribers and makes overages non-refundable.
A worked example
A Shopify store handling 1,000 support conversations a month:
Assumption | Value |
Support conversations per month | 1,000 |
Zipchat replies needed (at ~2.4 per conversation) | ~2,400 |
Fitting Zipchat tier | Pro, $249/mo (3,000 replies) |
Effective Zipchat cost per conversation | ~$0.25 |
Zipchat overage rate past 3,000 replies | $0.196 per reply |
My AskAI, helpdesk chat at ~$0.10 per ticket | ~$100 of credit |
The units differ, so read that as directional. Zipchat meters outbound AI messages on a storefront widget; My AskAI meters tickets inside your helpdesk. Both sit on top of the helpdesk seats you pay for either way, so compare AI charge to AI charge.
My AskAI is a flat ~$0.10 a ticket on Pro at $199, Scale at $499 or Enterprise from $999, with no separate reply meter. The ROI calculator will run your own volume. And you can prove it before you pay: 30 days free, all features unlocked, unlimited tickets, no card.
Does Zipchat have a free trial?
TL;DR: Seven days on every plan, with a card required up front. The more interesting free option is the permanent free plan on the Shopify App Store, which the vendor's own site never mentions.
The pricing page offers a "7-day free trial on every plan" alongside a "30-day money-back guarantee". Starting the trial means picking a plan and entering billing details, or going through Shopify's billing flow. It isn't card-free, and no extension is documented.
At the end, the trial converts to whichever plan you picked and the 30-day refund window starts, so I'd put that date in the calendar. That window is monthly-only, so an annual subscriber taking the 20% discount gives up the refund path.
Demos stay optional either way, and only the Unlimited tier needs a sales conversation.
The better free option is the plan on the Shopify App Store. I'd spend that allowance before the trial clock starts.
You get 120 AI replies a month, 100 training pages, unlimited human replies, corrections, product recommendations, order status and the full channel set, permanently. That's about 50 conversations a month, enough to judge answer quality on your own catalog before you spend anything.
My AskAI's trial is 30 days with every feature unlocked, unlimited tickets and no credit card. New customers are usually offered 50% off the first three months.
Is Zipchat worth it?
TL;DR: Worth it if your problem is unanswered pre-purchase questions on a storefront. Not worth it if your problem is ticket volume and cost per ticket in a helpdesk.
Zipchat is well built, ships fast and is fairly priced. The places it looks weakest on paper (reporting, resolution definitions, helpdesk breadth, knowledge connectors) all follow from building a storefront sales agent. I don't read any of them as mistakes.
Choose Zipchat if:
- Unanswered pre-purchase questions are costing you sales, and you want recommendation, proactive chat and cart recovery built in as the product.
- You want one agent across web chat, WhatsApp, Instagram DMs, Messenger and email, with 95+ languages and no per-channel pricing.
- You want to be live the same day with no sales call, at a low price with nothing feature-gated and unlimited seats.
- Your monthly volume sits comfortably inside a tier, so the 2x overage never bites.
Don't choose Zipchat if:
- You're a support-led team measured on resolution and cost per ticket. The meter charges support chatter at the sale rate, the dashboard reports revenue, and the published rates run 70% to 97% with no definition attached.
- Your support runs on Gorgias, Intercom, Freshdesk, HubSpot, Help Scout or Kustomer, so Zipchat sits beside your ticketing and never inside it.
- Your knowledge lives in a wiki or a cloud drive, needs to bind to live documents, or should come from your own past tickets.
- You need voice or SMS, or you need SOC 2, ISO 27001, SSO, EU data residency or an SLA to clear a security review.
How does Zipchat compare with My AskAI?
The criteria buyers shortlist on:
Criteria | Zipchat | My AskAI |
Where the agent runs | Storefront widget on your site | Inside the helpdesk you run |
Helpdesks covered | Zendesk only | Zendesk, Intercom, Freshdesk, Gorgias, HubSpot |
Pricing model | Metered per AI reply | Flat rate per ticket, ~$0.10 |
Published handling rate | Four figures, none defined | 72%+ on a rolling 30 days |
Knowledge sources | Your URLs, files and pasted text | Help center, site, files, cloud drives |
Security evidence | GDPR badge, self-declared | SOC 2 Type II, GDPR, live trust center |
Languages | 95+, auto-detected | 95, auto-detected |
Free trial | 7 days, card up front | 30 days, all features, no card |
Most of what a shopper asks before buying is a support question wearing a sales hat. Does it come in my size, is it in stock, when does it land, will it work with what I already own?
We lead with pre-sales support for exactly those questions, and a Shopify Task covers the recommendations.
The pure shopping-assistant vendors have built a long way down the conversion path, with product cards, code application and revenue attribution. We sell one agent doing both jobs, inside whichever of the five helpdesks you run (Zendesk, Intercom, Freshdesk, Gorgias or HubSpot), and we start from getting the pre-purchase answer right.
For the wider field: the best AI chatbots for ecommerce pre-sales roundup covers the sales-led group, and the Yuma AI and Gorgias Automate alternatives posts cover the helpdesk-native side.
What are the pros and cons of Zipchat?
Pros
- Purpose-built for the sales side of store chat. Recommendation, proactive chat, campaigns and single-use discount codes are the product itself, and catalog-level knowledge is the most praised thing about it (see features).
- Shopify-native install, live off the catalog in minutes. Six documented steps, an automatic crawl on install, and a free onboarding call on every plan (see setup).
- Flat capability at a low price. Every AI feature, five channels, 95+ languages and unlimited seats from $49, plus a permanent free plan through Shopify (see channels and cost).
Cons
- It optimizes conversion, so your support metrics come second. The ten dashboard tiles carry no CSAT, first-response time, resolution or deflection metric, and Good Replies counts unrated replies as successes (see improving responses).
- Four published handling rates under three labels. All four share a single attribution with no methodology behind it, and the worked example doesn't reconcile with its own escalation figure (see limitations).
- Metered per AI reply, at double once you overrun. A conversation is roughly 2.4 replies, so support chatter burns the same allowance as a sale (see cost).
- One helpdesk integration, and it's Zendesk. Any other ticketing platform means running Zipchat beside your inbox (see channels).
Zipchat
- Brand: ZIPCHAT INC
- Rating: 7/10
- In a sentence: A very good AI agent for the problem it was built to solve, which is selling, as long as you've checked that selling is the problem you have.
FAQs
How much does Zipchat cost?
From $49 a month on Starter (500 AI replies) to $499 on Scale, with Unlimited from $999 behind a demo. Growth sits at $129 and Pro at $249, and I'd size the tier off your busiest month. There's also a permanent free plan at 120 AI replies a month, but only through the Shopify App Store, and overage runs $49 per extra 250 replies.
Does Zipchat work with Shopify?
Yes, and Shopify is where it's deepest. It's a certified app carrying the Built for Shopify badge, installed through a theme app embed, and it syncs your product catalog and order data directly. WooCommerce, BigCommerce, Magento, PrestaShop and Wix are supported as well, though I'd expect less depth on each.
Is Zipchat worth it?
It depends which problem you're solving, so get that straight before you shortlist anything. If unanswered pre-purchase questions are costing you sales, it's purpose-built for that and fairly priced. If you're measured on resolution and cost per ticket, the reply meter and the revenue-first reporting will fight you.
What is Zipchat AI?
A standalone AI chat agent for ecommerce storefronts, and I'd read the emphasis on selling as deliberate. It recommends products, recovers carts and answers support questions across web chat, WhatsApp, Instagram DMs, Messenger and email. It isn't a helpdesk, and other than Zendesk it doesn't run inside one.
Does Zipchat have a free trial?
Seven days on every plan, with a payment method required to start, so set a calendar reminder. There's also the Shopify App Store free plan, at 120 AI replies a month, permanently. A 30-day money-back guarantee sits behind that, restricted by the terms to monthly subscribers.
What channels does Zipchat support?
Website chat (a bubble, a standalone chat page or an embedded iframe), WhatsApp, Instagram DMs, Facebook Messenger and email, all on every paid tier with no gating. There's no voice channel and no first-party SMS.
Can Zipchat handle refunds and returns?
It answers returns and refund questions from your policies. Return or exchange workflows run as an agentic skill you build against your own API, webhook or MCP server. The pre-built tool set is narrower: knowledge search, escalate, order lookup, create discounts on Shopify, and skip conversation.
Can Zipchat look up an order?
Yes. Order lookup is one of the five documented runtime tools, native on Shopify and available elsewhere through a custom order-tracking API. It's the first thing I'd test on a trial.
Does Zipchat hand over to a human agent?
Yes. Escalation is a documented tool, your team can take over mid-conversation from the inbox, and on Zendesk the handover lands in the same thread. Zipchat claims under 3% of conversations escalate, the figure that doesn't reconcile with its own worked example.
What AI model does Zipchat use?
It doesn't say, and it says so explicitly. The AI platform page reads "Zipchat is built on leading frontier AI models", then "We do not name specific models publicly." Its subprocessor list names AWS and Google Cloud but no AI vendor, so treat the model as unknown.
What are the best Zipchat alternatives?
It depends which limit you hit. Tickets in Gorgias, Intercom, Freshdesk or HubSpot mean you want an agent that runs inside that helpdesk, and wanting deeper conversion actions means you want another shopping-assistant vendor. The best AI chatbots for ecommerce pre-sales roundup covers the second group and the Yuma AI alternatives post covers the first.
