<!-- Generated from vs/ai-chatbots.html by scripts/generate-md-twins.mjs - do not edit by hand. -->

# AI concierge vs AI chatbots

A generic AI chatbot answers inside a chat bubble, drawing on broad general knowledge, which makes it good at open-ended conversation and bad at being certain about your business. An [AI concierge](https://askmywebsiteai.com/what-is-an-ai-concierge) is deliberately narrower: it answers only from the website's own content plus notes the owner adds, it reads the visible text of the page the visitor is currently on, and instead of describing where to click it highlights the element, step by step, across pages. The trade is scope for reliability, and description for arrival.

## Where generic chatbots are genuinely better

The honest version first, because the categories are good at different jobs.

-   **World knowledge.** A general model can explain a legal concept, translate an idiom, compare industry approaches or summarise a field. A concierge grounded in one website cannot, and should not pretend to.
-   **Open-ended conversation.** Brainstorming, drafting, reasoning through a problem that has nothing to do with what is on the page.
-   **Breadth outside the site.** If your visitors genuinely need answers your website does not contain, a general assistant covers more ground.

If what you want on your site is a general-purpose conversational partner, a broad model is the right tool. If what you want is for visitors to stop bouncing because they could not find the shipping policy or the plan-change button, breadth is not the problem you have.

## Grounding: where the answer comes from

The single most consequential difference is the source of the answer. A general model produces plausible text about your business from whatever it absorbed, which is how a chat widget ends up inventing a return window, a phone number, or a feature you do not ship. Plausible and wrong is worse than nothing, because the visitor believes it and you find out at the complaint stage.

AskMyWebsite AI scans the site when it is set up, and answers from that content plus the owner's notes, and from nothing else. It does not consult the open internet. That constraint is not a limitation grudgingly accepted; it is the product. It means the answers a visitor gets are answers you published, and when the content does not cover something, saying so is the correct outcome rather than a failed one.

## Context: what the assistant can see

Most chat widgets see only the message the visitor typed. The visitor, meanwhile, is looking at a page full of relevant information and reasonably assumes the assistant is looking at it too. So they write "is this one waterproof?" and get asked which product they mean.

The concierge reads the visible text of the current page, so "the green chair" is unambiguous, "this plan" resolves, and "what does this field mean?" has an antecedent. It reads visible text only - values typed into form fields are never captured, anywhere in the flow.

## Guidance: describing versus arriving

A chatbot's answer to "how do I cancel my subscription?" is a paragraph of directions. The visitor then has to translate that paragraph into your actual interface, which may have been redesigned since whoever wrote the help article last looked at it.

The concierge answers and then highlights the exact element to click, one step at a time, continuing across page navigations until the visitor is where they were going. And it stops there: it never clicks, fills or submits anything itself. The visitor always acts. That is a security and consent property rather than a missing feature - an assistant with no ability to act cannot be persuaded by anything on a page to act wrongly, and decisions like buying or cancelling stay with the person making them.

It also replies in the language the visitor writes in, regardless of the language your site is published in.

## Pricing: pay for answers, not tokens

Token-based pricing charges for the volume of text the model reads and writes. From the buyer's side that is an odd unit: you are billed for the assistant's verbosity and for how much context the vendor decided to send, neither of which you control, and none of which maps to anything you would put on a business plan. A long-winded model costs more than a concise one while delivering the same outcome.

AskMyWebsite AI charges for answers, not tokens. One answer means one visitor question answered. The guidance continuations after an answer, as the assistant walks someone through the remaining steps, are free - so a five-step flow is one answer, not five. When an account reaches its cap, visitors already mid-conversation finish their thread, and for new visitors the widget simply does not appear. Your site never shows a visitor an error.

AskMyWebsite AI is in private beta and we are onboarding a few websites at a time. Tell us your website address and what you would want the concierge to do, through the [contact page](https://askmywebsiteai.com/contact), and we will come back to you with access and pricing.

## Deployment and privacy

-   **Installation:** one script tag before `</body>`. No plugin, no changes to your site's code. It runs on WordPress, Shopify, Webflow, Squarespace, Wix, Framer and custom HTML, and it detects client-side route changes so single-page applications behave correctly.
-   **Data:** EU operator (Spain), EU hosting (Barcelona), a GDPR Article 28 data processing agreement at [/dpa](https://askmywebsiteai.com/dpa).
-   **Cookies:** the widget sets none. It uses sessionStorage only after the visitor opens it, and that dies with the tab.
-   **Retention:** conversations are auto-deleted after 90 days, and your content is not used to train AI models.
-   **Payments:** only on Stripe hosted pages.

## Choosing between them

Ask what a good outcome looks like on your site. If it is "the visitor had an interesting conversation", breadth wins. If it is "the visitor got a correct answer about our business and then completed the thing they came to do", then grounding, page context and on-page guidance are the three properties that produce it, and general knowledge is beside the point.

Nothing stops a business from using both for different purposes. But on the website itself, the narrower assistant is usually the one that moves the numbers you care about, because the reason people leave is rarely that your widget lacked opinions about the wider world.

## See it on your own pages

The test that settles it takes a few minutes: install it, ask the question your customers ask most, and see whether the answer is yours.

[Get in touch](https://askmywebsiteai.com/contact)

AskMyWebsite AI is in private beta. Tell us about your site through the [contact page](https://askmywebsiteai.com/contact), and we will come back to you with access and pricing. You can also look around from [the homepage](https://askmywebsiteai.com/).

## Related

-   [What is an AI concierge](https://askmywebsiteai.com/what-is-an-ai-concierge)
-   [AI concierge vs live chat](https://askmywebsiteai.com/vs/live-chat)
-   [AI concierge vs AI chatbots](https://askmywebsiteai.com/vs/ai-chatbots)
-   [AI concierge vs site search](https://askmywebsiteai.com/vs/site-search)

---

Canonical page: https://askmywebsiteai.com/vs/ai-chatbots
Site index for AI systems: https://askmywebsiteai.com/llms.txt
