For most business sites, not yet. A chatbot earns its place when you handle a high volume of repetitive questions and have solid documentation behind it. For a marketing site with a dozen pages and a contact form, it usually adds cost and friction rather than removing it.
This comes up in almost every kickoff now. Someone has seen a competitor's chat bubble and wants one, and the request arrives as a feature rather than as a problem to solve.
Our answer depends entirely on what the site is for. Here is how we think it through, including the cases where we say yes.
It answers visitor questions in conversation, usually by drawing on your existing content. Modern versions use a language model grounded in your documentation, so they can answer things nobody scripted. Older rule-based bots follow fixed decision trees and break the moment someone asks something unexpected.
The distinction matters commercially. A scripted bot is a menu wearing a conversational costume. A grounded model can genuinely handle variation, but it can also produce a confident wrong answer, which a decision tree never does.
Both share one dependency. A chatbot can only be as good as the material behind it. If your documentation is thin or out of date, the bot inherits that problem and repeats it to every visitor.
More than they were, and less than the hype suggests. A Pew Research Center survey of 5,123 US adults, conducted in early 2025, found that 34% had used ChatGPT, roughly double the share in 2023. It also found that 66% had never used it at all.
Age splits it sharply. The same Pew survey found 58% of adults under 30 had used ChatGPT. If your customers are older, a conversational interface is less familiar ground than it might feel from inside a tech company.
Trust in AI answers is rising in commercial contexts too. BrightLocal's Local Consumer Review Survey 2026 found that 45% of consumers now use AI tools for local business recommendations, up from 6% the previous year, and that 40% trust AI platforms to provide those recommendations.
The honest read is that comfort is growing fast but is far from universal. Designing as if everyone is fluent will lose you the third of visitors who are not.
When question volume is high, answers are documented, and the alternative is a slow human queue. Support portals, software products, and businesses with complex policies all qualify. In those settings a good bot deflects real work and answers at three in the morning.
The clearest signal is repetition. If your team answers the same fifteen questions every week, that is a job a well-grounded bot can take on, and you already have the raw material to build it.
Scale matters as well. A bot handling forty conversations a day pays for itself. A bot handling two conversations a week costs more in maintenance and monitoring than it saves.
Complexity is the third case. When customers genuinely need to navigate options, eligibility rules, or specifications, guided conversation beats making them read a long page and work it out.
When it stands between a motivated buyer and a human. If someone is ready to talk about a project and gets a bot asking qualifying questions, you have added friction at the exact moment you should be removing it. That is the most expensive place to put one.
Wrong answers are the other risk, and they are worse than no answer. A model that confidently states an incorrect price, policy, or capability creates a commitment you then have to walk back with a real customer.
The third failure is quieter. A bot placed on a site with thin content becomes a substitute for writing the pages properly. The information still does not exist, it is just now hidden behind a conversation.
We are blunt with clients about this one. If the answer is not on your site, a chatbot does not solve that. It disguises it.
Third-party chat widgets are among the heavier scripts a site can carry, and they typically load on every page whether anyone opens them or not. On a site that has been carefully optimised, one widget can undo a meaningful share of that work.
The cost is not only bytes. These scripts run on the main thread, which is the same resource the browser needs for responding to taps and clicks. That shows up in interaction responsiveness rather than in load time alone.
Across our own builds the average PageSpeed score is 98, and third-party embeds are the single most common reason a score like that starts slipping after handover. We wrote about that pattern in our piece on how third-party scripts slow down your website.
There are ways to soften it. Loading the widget only after user interaction, or only on the pages where support questions actually arrive, keeps most of the benefit and much less of the cost.
It does very little either way, and that is the point people miss. Content locked inside a chatbot is invisible to search engines and AI answer engines. The same answers published as pages can be crawled, indexed, and cited.
This is the strategic argument against chat-first thinking. If you have written good answers to fifteen common questions, putting them only in a bot means they can never earn you a search result or a citation in an AI answer.
Published pages work harder. They serve the visitor who prefers to read, they can be found by someone searching, and they can be quoted by an assistant answering on someone else's screen.
Machine readable structure matters more than a conversational layer here, which is the argument we made in our piece on whether AI agents can use your website.
Far less widely than the industry conversation implies. The HTTP Archive's 2025 Web Almanac measured browser-based AI APIs on only a tiny fraction of the web, with the Prompt API detected on 0.095% of all desktop sites and 0.078% of all mobile sites it crawled that year.
Preparation for AI readers is similarly thin. The same 2025 Web Almanac chapter found that only 2.13% of desktop sites had valid llms.txt entries, while 94.1% of sites had robots.txt directives in place.
The naming gold rush is more visible than the implementation. The same chapter counted 11,848 .ai domains in the top 10 million sites in 2025, up from 2,824 in 2022. Registering the domain has been easier than shipping the feature.
We read this as a reason to be deliberate rather than late. Very few sites have done the groundwork, so the advantage is available to whoever does it properly rather than fastest.
Very often, yes. Most of what businesses want from a chatbot is faster answers and more enquiries. A clear contact page, honest pricing information, and a genuinely good FAQ deliver both without a subscription or a script on every page.
The comparison is rarely made fairly. A chatbot is judged against doing nothing, when it should be judged against spending the same money on the content the bot would have needed anyway.
Response time is the part customers actually feel. We reply to enquiries within 48 hours, and in our experience knowing when you will hear back matters more to people than getting an instant automated reply that resolves nothing.
We went through what makes that page work in our post on designing a contact page that gets real enquiries. It is a cheaper experiment than a chatbot and it is reversible.
Add one if you handle high volumes of repeat questions, have documentation worth grounding it in, and can monitor what it tells people. Skip it if your site is small, your content is thin, or your main goal is getting serious buyers to a conversation with a person.
If you do add one, ground it in your own content, publish those answers as real pages too, load it only where it is needed, and give every conversation an obvious route to a human.
Our overall position is that a chatbot is an amplifier rather than a fix. It makes good content easier to reach and makes thin content fail faster and more publicly.
If you are weighing this up and want a straight opinion on whether it is worth it for your site, we are happy to talk it through. Reach out via phoenix.studio and we will get back to you within 48 hours.
Tell us where you want to go. We'll tell you how we'd get you there.