Comparison
Refik and Gladly, read on the same day.
One of these two prints what it costs on a page. For the other we went looking for that page and did not find it: on the day of reading, the address that should have held it answered with the home page instead. That is the first honest sentence here, and it is a statement about our reading rather than about their product — so most of the right-hand column below is empty, and every empty cell says why underneath.
Method
Two dates, an address that moved, and a column that is mostly empty.
The right-hand side of every row was read out of Gladly’s own pages on 10 October 2026 — three fetches, and only two distinct documents, because one of the three came back byte for byte identical to another. The quotes are theirs, word for word. No analyst report and no review site was used, and no integration directory or sales conversation was part of it. The left-hand side was read out of our own source code on 4 October 2026, and the plan figures come from the pricing page in this same directory, which fails its build if a price in the code changes without the page changing with it.
There is a wrinkle in the addresses and it belongs at the top rather than in a footnote. Gladly’s pages have moved: gladly.com now answers from gladly.ai. Both addresses were used on the day of reading, and every quote in the table says which one it came from. A quote carrying neither a date nor an address, taken from a site that has just moved, is a claim with nothing behind it.
An empty cell is a third state and it is on purpose. It does not mean “they do not have it”. It means we did not read it on that day, so printing a dash there would turn a gap in our reading into a claim about their product. Seven of the twelve cells in their column are empty, which is the shape three fetches of a marketing site produces when you ask it twelve questions — and we would rather show that than fill the gaps with a guess. Where a cell is empty, on either side, the reason is written underneath the table.
Three things are missing from this page by design. There is no verdict row and no score: a printed band price and a price we could not find are not two values of the same quantity, and putting a number between them would produce something that looks like a measurement and is not one. There are no traffic, keyword or authority figures, because the tools that produce those were not connected on the day the pages were read, and an estimate would be an invention wearing a number. And there is nothing here about how well either assistant answers a shopper, because that was not measured on either side.
The rows
What we charge, and what their pages said.
Read the first two groups together. The first is our price, printed in full. The second is the search for theirs, written out as what was requested and what came back, because that is the only honest form a failed reading can take.
Gladly read from gladly.ai, gladly.ai/product/ai-customer-service-agent/, gladly.com/pricing/ and gladly.ai/sitemap.xml; Refik read from this repository and the pricing page.
| What the row measures | Refik | Gladly |
|---|---|---|
| What you are billed for | ||
| The billing unit | Conversations in a billing period. Three bands: 300, 1,000 and 3,000, for the whole shop rather than per person answering. | |
| Published list prices in USD | Starter $39 a month or $390 a year; then $99 and $249 a month, or $990 and $2,490 a year. | |
| Charges on top of the subscription | On monthly billing, $0.15, $0.12 or $0.10 for a conversation above the band, capped at one band’s worth. On yearly billing, none. | |
| Free trial | 14 days on every plan, with a conversation allowance scaled to the plan. | |
| Where the price was looked for | ||
| The address the pages were read from | This repository, and the pricing page in this same directory. | gladly.ai. The older gladly.com has moved there, and that move is why the row below matters: the old pricing address still answers, but not with a pricing page. |
| What a request for a pricing page returned | The pricing page, with the figures that are in the code. If a price changes in one place and not the other, the build stops. | A request for gladly.com/pricing/ ended on the gladly.ai home page, and that fetch was byte for byte the same file as the home page fetch. The sitemap at gladly.ai/sitemap.xml answered with 164 KB and listed no pricing page; the only entries carrying the word were blog posts. Across the pages read, the word “pricing” appeared no times in the visible text — and the same reader found the word “demo” six times, so it was not blind. |
| Where it runs | ||
| Platforms it connects to | Shopify only. The catalogue, the policies and the order lookups all come from one store’s Admin API, and there is no second platform in the code. | On the pages read that day, the word “Shopify” appeared no times in the visible text. The same reader found the word “Gladly” on every one of them, so it was not blind. This is a measurement of three fetches of a marketing site, not of their product. |
| Channels a shopper can use | The web storefront, and only that. WhatsApp, Instagram, Messenger, email and SMS are not built, and that is written on the integrations page rather than left to be discovered. | |
| What the assistant can be taught from | The shop’s catalogue, its published policies, and question-and-answer pairs typed in by hand up to a ceiling of three hundred. There is no URL crawling, no PDF or document upload, and no fetching of the shop’s other pages or its blog. | |
| A ticket in your help desk | The server can open a ticket in exactly one help desk, and it is not Gladly. The settings screen for choosing a target is not built either, so a shop cannot wire the handover up from our panel today. | |
| What each side calls itself | ||
| How each side describes itself | A shopping assistant on the storefront: it recommends, it looks up an order, and it hands a hard question to a person. | The home page is titled “Agentic commerce platform built for top brands | Gladly”, its description reads “Gladly brings agentic commerce to your site for premium ecommerce support. See how it works.”, and the line at the top of that page is “Commerce AI for every customer moment.” |
| What the product page is titled | “…agent: devotion, not deflection | Gladly”, at gladly.ai/product/ai-customer-service-agent/. | |
- The four price rows — the billing unit, published list prices, charges on top of the subscription, and the free trial — are all empty on their side. No publicly readable pricing page was found in two attempts: the pricing address itself, and the sitemap. That is a failed reading, not a finding about their product. Their tiers and amounts may be published somewhere we did not look, or given in a conversation; either is ordinary, and neither is something we can state. Prices are also the most volatile fact on a page like this, which is why ours carry the band they belong to and theirs carry nothing at all.
- Channels, what the assistant can be taught from, and the help-desk handover — their three cells are empty. None of the three pages fetched answered those questions, and an integration directory was not among what was read. Ours are written out because a missing channel and a missing teaching path are the kind of thing a shop discovers after it has paid.
- What the product page is titled — our cell is empty. The 4 October measurement read the product’s source code, not the titles on our own marketing pages, so there is no dated reading to put in that cell.
- The Shopify row measures their pages, not their product. A word appearing no times across three fetches of a marketing site says what those three documents contain. It cannot be read as “they do not support Shopify”, and we are not reading it that way.
- Neither side’s answer quality is here. Not theirs, not ours. Measuring it would need the same shoppers asking the same questions of both, on a real catalogue, and that has not been done.
- No traffic, keyword or authority numbers anywhere on this page. The tools that produce them were not connected on the day the pages were read. An estimate would look like a measurement and would be an invention.
The difference that matters
A price you can read, and a price you have to ask for.
The thing that separated these two on the day they were read is not a feature. It is that one of them answers “what does this cost” on a page, and for the other we could not find that page. Our three bands and their amounts are printed, the charge for a conversation above a band is printed, and the one case where that charge does not apply is printed. A shop can work out roughly what it will owe before it installs anything, and can check our arithmetic against the same figures the code uses.
On their side the honest sentence is shorter: the address that should hold a price answered with the home page, byte for byte, and the sitemap did not list one. That is a reading of two requests, and it has at least two innocent explanations — the page may live somewhere a sitemap does not reach, or the number may be given in a conversation rather than printed. Both are ordinary in this market. What a shop comparing the two will actually meet is the difference worth stating: on one side a figure is on a page, on the other the first step is to ask.
The second structural fact from the same reading is the moved address, and it is worth saying plainly because it is exactly the kind of thing that turns a comparison false while it sits there looking fine. gladly.com now answers from gladly.ai. A page that quoted the old address without saying so would be citing something that no longer serves those words, and a reader could not tell. Every quote above therefore carries the address it was read from, alongside the date, and both are there so that you can go and check whether either has changed since.
There is one more thing a shop on their platform should know before reading any further, and it is about us rather than them. Our server can open a ticket in exactly one help desk, that one is not Gladly, and the screen for choosing which help desk to target is not built. So if your support already lives inside Gladly, our handover cannot be wired into it from our panel as things stand; a shopper’s hard question would be handed to a person through our own inbox instead of arriving where your team already works. The integrations page lists what is connected and what is not, in the same words.
Who each one fits
Where we would tell you to go and ask them.
Go and get their price if the words they use for themselves are the words you would use for your problem: a platform for a brand’s whole relationship with its customers, across the places that relationship happens, rather than one storefront’s product pages. We cannot tell you what that costs, because we could not find it, and we are not going to guess at it in order to look better next to it. Ask them, and then compare what they tell you against the three bands printed above — that comparison is yours to make with two real figures, not ours to make with one.
We fit you if the questions that cost you money arrive before the order exists: which of these two is right for me, is this in stock in my size, where is the order I placed on Friday, can I send this back. Those are storefront questions, they arrive on a product page, and answering them needs the catalogue and the policies of one shop. We also fit you if you want the number before the conversation — if a published band and a published overage charge are worth more to you than a tailored quote.
Neither of us fits you if what you actually need is a person. Both products put a human at the end of the path; ours does it by opening a handover and telling the shopper that a person will pick it up, and it is told never to say a refund is approved. If there is nobody to pick it up, no assistant on either side of this table fixes that. And if your support lives in more than one inbox with more than one person answering, read the help-desk row again before you read anything else on this page.