What it does
CustomGPT.ai sells one thing well: a hosted RAG pipeline you rent instead of build, where citations are on by default. Point it at a website, a sitemap, or a document pile and it returns an embeddable chatbot that answers and shows the source URLs it pulled from. The vendor owns ingest, retrieval, and the model call. That is the entire value proposition, and it is a real one, because the part teams underestimate when they build this themselves is not the model call, it is keeping retrieval fresh and citations correct as the source docs change. The buyer it fits is the SMB support team or internal-knowledge owner who wants grounded answers and explicitly does not want to operate a vector store.
What you get
Ingestion is broad. URL crawl, sitemap, and direct file uploads (PDF, DOCX, XLSX, CSV, JSON, MD, PPTX), plus connectors for SharePoint, OneDrive, Confluence, Google Drive, Vimeo, and YouTube. The detail that earns its keep in a support context: citations are on by default. Every answer can carry the source URLs it drew from, which is the single biggest lever against a confidently wrong reply landing in front of a customer.
Three ways to ship the bot. Embed it on your domain, query it from your own app over the REST API, or pipe events into Zapier. More than 90 languages work across every paid tier. White-labeling and auto-sync are gated to Premium and up.
Who should use it
The fit is a team with somewhere between 50 and 5,000 documents and a small ops crew that wants a managed pipeline rather than a self-hosted one. SMB customer service, internal knowledge bots, docs-site assistants: those are the natural homes. Skip it if you want to self-host, if you need code-level control over the agent loop, or if your budget tops out below 100 dollars a month. None of those are what this product is for.
Pricing
Standard runs 99 dollars per month, or 89 per month billed annually, and covers 10 agents and 1,000 queries a month. Premium jumps to 499 per month (449 annual) for 25 agents and 5,000 queries. Enterprise is custom-priced, unlimited on both. There is a 7-day free trial with full Standard features. A credit card is required at signup, so set a calendar reminder before day seven if you are only kicking the tires.
The 1,000-query Standard cap is the number that decides fit, not the price. A docs-site assistant or internal bot for a team of 50 stays well under it; a public-facing support bot on a site with real traffic will blow through 1,000 queries in days, and at that point the per-query economics of Premium need to clear against a self-hosted retrieval stack you actually staff. The query ceiling, not the dollar figure, is where this product stops being the obvious answer.
Verdict
Buy it when the job is "ground a chatbot in our own docs, with citations, and never think about a vector store again," your document set is in the 50-to-5,000 range, and your query volume is internal-scale rather than public-internet-scale. The citation-by-default behavior is the single feature that earns the subscription, because it is the one thing that keeps a confidently wrong answer from reaching a customer unflagged. The recommendation flips the moment you need code-level control of the agent loop, want to self-host, or your query volume crosses into territory where renting the pipeline costs more than owning a thin one. Rated 4.0: a sharp, well-scoped product that is genuinely the right call inside its lane and genuinely the wrong one outside it.