Chatbase vs CustomGPT, compared on price and made-up answers

Chatbase charges in message credits, and you can start on its free plan, with 50 a month. A reply costs 1 to 6 credits depending on the model, and an action can double that with a second request. Credits renew on the 1st of each calendar month, and unused ones don’t carry over. Hobby, at $40 a month, or $32 a month billed yearly, has 700 credits, which is 116 to 700 replies, or fewer when actions run. An agent on the free plan is deleted after 14 days without use. Chatbase pricing sets out what happens when the credits run out.

CustomGPT charges in queries, and there’s no free plan. The 7-day trial of Standard needs a card, and when it ends you’re charged for the plan you chose, unless you cancel. Standard is $99 a month, or $89 a month billed yearly, for 1,000 queries, 10 agents, 3 team members and 60 million words of content. Every message your agents process counts as a query, visitors’ messages included, and actions such as Lead Capture cost extra queries. All your agents draw on the same balance. Premium, at $499 a month, or $449 billed yearly, has 5,000 queries and 25 agents, and removes the CustomGPT branding.

The gap widens when you run out. Chatbase’s auto recharge, if you’ve switched it on, adds credits at $40 per 1,000. On CustomGPT, the agent stops responding until the next cycle, and the add-on that takes you past the limit is 2,500 extra queries at $375 a month, more than three times the price of Standard itself. So Chatbase is cheaper to start and to top up, and CustomGPT sells room, with more agents and far more content, than a single product usually needs.

Choose Chatbase if you want to start free and be answering visitors within the hour. The same agent runs on 16 channels, can let a customer track a Shopify order or change their Stripe plan, and grows into a helpdesk and phone calls on Standard, at $150 a month. Just write the instruction to stay inside your content yourself, because Chatbase leaves that to you.

Choose CustomGPT if an answer outside your content would cost you dearly, and you want citations and a boundary around your content built in. It also suits a large archive, with 60 million words on its first plan, over 1,400 file formats and YouTube and audio as sources, and an agency running agents for several clients. Budget for the card-first trial and the price of going over.

Neither is shaped for a founder who answers support alone. One makes you watch a credit meter, and the other charges for room you won’t use. There’s a third option, built for exactly that founder.

The fear with any support bot is that it invents a refund policy in front of your customer. Deacon answers only from what it finds in your own docs, and when they don’t cover a question, it says it doesn’t know and asks for the visitor’s email. There’s no instruction for you to write first. It answers in seconds, at any hour, in the language your customer writes in, and an answer taken from a page on your site usually links to that page. You get an email when a visitor leaves an address or asks for a person, and you reply from Deacon’s dashboard.

The questions it can’t answer are where it improves. Each one your docs didn’t cover gets an Answer this button, and the answer you write becomes part of what Deacon knows. Save & check shows you the reply Deacon now gives and how many other open questions your answer covers, so it can answer the next person who asks, and it needs you less each time.

You also learn what your customers need. Every conversation is saved and searchable by what visitors asked. Topics group similar questions, count the people behind them and label most of them, as a feature ask, a bug report or a how-to, so the most-asked question is at the top of your list. The same script runs website analytics too, with visitors, sources, goals and funnels, on every plan.

And trying it costs nothing. The free plan needs no card, has no end date and deletes nothing for inactivity. Every feature is on every plan, and the plans differ only in answers, knowledge, pages from your website and seats. Every plan answers with the same model, so no credits run out faster on a better one. Nothing is billed past your plan’s answers, no plan upgrades itself, and there’s no badge to pay to remove.