SiteGPT pricing, and what a message really buys

SiteGPT pricing is per plan, and there’s no free plan. Starter is $59 a month, Growth $129 and Scale $429, or $39, $79 and $259 a month billed yearly, and Enterprise is priced after a demo. Every self-serve plan starts with a 7-day free trial that needs a card. We read every figure on SiteGPT’s pricing page and its docs on 25 September 2026.

The allowance is what to understand before you pay. Starter offers up to 4,000 messages, but SiteGPT’s docs count the visitor’s question and the chatbot’s reply as a message each. A new chatbot also answers with GPT-4.1, whose messages cost ten times as much. With the whole allowance on GPT-4.1, Starter covers 200 questions answered a month.

Deacon is ours. It’s an AI customer support agent for founders who answer support themselves, and it counts only the replies it writes, with the same model on every plan. The free plan gives you 50 answers a month, with no card and no end date.

SiteGPT’s plans side by side

The pricing page opens on yearly billing and says you save up to 40%. By our arithmetic the saving is 34% on Starter, 39% on Growth and 40% on Scale.

The messages are a monthly allowance, and the limits are from SiteGPT’s plans and limits page. A page of content is 2,500 characters of text, counted per document and rounded up, so a 5,000-character file is 2 pages and a 5,001-character one is 3. Your pages are shared across your chatbots. Prices exclude taxes.

Each step up buys more than messages. Growth adds the integrations with Slack, Messenger, Google Chat, Crisp chat, Freshdesk and Zoho, API access, the iPhone app, a monthly sync of your content, a cap on messages per conversation, and analytics with daily trends, CSV export, a weekly digest and a list of unanswered questions. Scale adds the Zendesk Messaging integration, webhooks, a weekly sync, a daily scan of your sitemap and AI insights that sort conversations by topic, sentiment and outcome. On Starter you refresh your content by hand.

What SiteGPT counts as a message

Both sides of a chat count. SiteGPT’s plans and limits page says “A visitor’s message is 1 message, and the chatbot’s reply is 1 message. So one question and its answer use 2 messages.” The pricing page’s own line, that each message “counts both the user’s question and the AI’s reply”, is easy to read the other way, so go by the docs.

The model is the second multiplier. A chatbot answers with GPT-4.1 mini or GPT-4.1, and one GPT-4.1 message uses as much of the allowance as 10 on mini. SiteGPT’s model guide says new chatbots use GPT-4.1, and advises switching most of them to mini. You split one allowance between the two models with a slider on the Usage page.

Put the two together and each plan answers far fewer questions than its headline figure. These counts are ours, halving each plan’s messages.

The allowance is shared by all your chatbots and resets on the 1st of each calendar month (UTC), whatever your billing date. Unused messages don’t carry over.

What happens when SiteGPT’s messages run out

When one model’s share is used up, the chatbot switches to the other while it has messages left. When both are gone, your chatbots keep answering visitors, on your site, in integrations and through the API. What stops is your dashboard. Every chatbot page is blurred behind a card that reads “Message Limit Reached!” and asks you to buy the Extra Messages add-on.

SiteGPT’s terms carry a fair usage policy too. Under it, going past the parameters SiteGPT sets “could lead to restricted or reduced service access, with or without prior warning”. So in practice you buy packs of 5,000 extra messages, at $59 a month each or $39 billed yearly, or you move up a plan.

Content has a cap of its own. At the page limit, new content is blocked, and documents past it aren’t trained until you delete something or upgrade.

SiteGPT’s add-ons

SiteGPT sells two add-ons, on any plan with a subscription.

The branding add-on costs as much as Starter itself, and no plan includes it. SiteGPT’s billing guide says add-on changes are charged or credited straight away, and not charged during a trial. The trial box on the pricing page also mentions adding webhook support from billing, but the billing guide lists only these two add-ons, and the plans table puts webhooks on Scale.

SiteGPT’s trial, and cancelling

The trial takes a card. SiteGPT’s terms say that on its last day, unless you’ve cancelled, you’re charged for the plan you chose. You get that plan’s full limits while it runs. From the billing page you can end the trial early and be billed at once, or extend it by 7 days from the cancellation dialog, once you’ve said in at least 25 characters what you need the time for.

A plan change starts at once, with the difference charged or credited right away. Cancelling a paid plan keeps your chatbots answering to the end of the period, while cancelling during the trial ends it immediately and the chatbots stop. Once a subscription ends, your limits drop to zero, and SiteGPT schedules your chatbots, conversations, documents and leads for deletion no earlier than 30 days later, emailing you the date at least 7 days before.

What a founder’s month costs on SiteGPT

Take a founder whose site has 300 conversations a month. Across 1,053 conversations Deacon answered, measured on 22 September 2026, a conversation ran to 1.95 replies, so call it 2.

That’s 600 questions and 600 replies, or 1,200 messages on SiteGPT’s meter. A busier month of 1,000 conversations comes to 4,000. Each bill below is the cheapest mix of plan and packs, worked out by us from SiteGPT’s pricing page, taking a pack’s 5,000 messages to count like the plan’s own, so 500 on GPT-4.1.

Taking the badge off adds $59 to each monthly bill and $39 to each yearly one.

The model decides the bill more than the traffic does. The same 1,000 conversations cost $59 a month on GPT-4.1 mini and $429 on GPT-4.1, and a chatbot you haven’t switched is on GPT-4.1. There’s a wider look at tools priced this way in how much an AI chatbot costs.

What the same month costs on Deacon

Deacon counts answers, and an answer is one reply Deacon writes. The visitor’s question isn’t counted, and every plan answers with the same model, so there’s no multiplier to choose.

The founder’s 600 replies fit in Starter’s 1,000 answers a month, at $50 a month or $40 a month billed yearly. The busier month’s 2,000 replies fit in Growth’s 4,000 answers a month, at $150 a month or $120 a month billed yearly. A reply in which Deacon says it doesn’t know counts as an answer, and chats you try in the Playground never count.

The free plan gives you 50 answers a month for one person, with no card and no end date, and nothing on it is deleted if you go quiet. When a month’s answers run out, nothing more is billed and no plan upgrades itself. Your visitors get a short reply saying Deacon can’t look into that one right now, and they’re asked for their email unless they’ve already left it in that conversation. Every owner on your team is emailed when 80% of the month’s answers are used, and again at 100%. There’s no “powered by” line on any plan.

Why founders who answer support pick Deacon

SiteGPT is built to answer from your content, with integrations for the tools a support team already uses. Deacon is built for the founder who answers the questions it can’t. It answers your customers in seconds from your website, files and pasted text, and when your content doesn’t cover a question it says it doesn’t know instead of guessing, and asks for the visitor’s email. You’re emailed when they leave it or ask for a person, and your reply from the dashboard appears in their chat while it’s open, on every plan.

Every question your content missed waits in the dashboard with an Answer this button. Write the answer and press Save & check, and Deacon asks itself the question again, shows you the reply it now gives and tells you how many other open questions your answer covers. So it gets better the more you answer.

SiteGPT’s list of unanswered questions starts on Growth, and its topic insights on Scale. Deacon’s are on the free plan. Topics group what visitors ask, count the people asking and label most groups as a how-to, a feature ask, a bug report and so on, and every conversation is saved and searchable by what visitors asked.

Questions about SiteGPT pricing

Does SiteGPT have a free plan?

No. The pricing page answers “No, there is no free plan.” Every self-serve plan has a 7-day trial instead, which needs a card and charges it on the last day unless you cancel. You can also try a demo bot on SiteGPT’s site, trained on its own pages.

How much is SiteGPT a year?

$468 for Starter, $948 for Growth and $3,108 for Scale, which is $39, $79 and $259 a month. Billed monthly, the same plans are $59, $129 and $429.

Do SiteGPT’s messages roll over?

No. The allowance resets on the 1st of each month (UTC), whatever your billing date, and unused messages don’t carry over. It’s shared by all the chatbots you own, and extra packs add 5,000 every month for as long as you keep them.

How do I remove “Powered by SiteGPT”?

Buy the Remove SiteGPT Branding add-on, at $59 a month or $39 a month billed yearly, on any plan. Then edit or remove the line in the widget’s Appearance settings. White labelling SiteGPT for your own clients is priced separately, by email.

Which model should I choose on SiteGPT?

SiteGPT’s own guide recommends GPT-4.1 mini for most chatbots, and GPT-4.1 only when answers don’t follow your instructions closely enough. Mini gets you ten times the messages, and a new chatbot starts on GPT-4.1, so check the setting before you go live.

Can Deacon replace SiteGPT?

If you want your customers answered from your own content, and to step in yourself when it can’t, yes. Deacon reads your website, files, pasted text and the answers you write. It has no integrations with Slack, Messenger or Zendesk, no webhooks, no public API beyond two install checks and no YouTube source, and your team replies from Deacon’s own dashboard. SiteGPT alternatives compares seven tools that start free, and Chatbase pricing prices SiteGPT’s nearest rival the same way.