Migration
Moving from SiteGPT to Creobot?
Keep SiteGPT live while Creobot proves the gap. We benchmark both assistants side by side on your own public pages, using your own questions, and hand you the source-grounded scorecard. Switch when it makes the case.
- Your current bot stays liveNothing is switched off
- Public pages firstNo sensitive data required
- Done-for-you benchmarkSource mapping and setup are on us
- Source-grounded scorecardEvery citation opened and checked
Fit
Who this is for
Teams move for handoff quality and lead signal, and they stay on SiteGPT because re-training on a new tool means re-doing source selection by hand.
You already run SiteGPT
This page is about switching cost, not about whether you need an assistant. You have one.
Answer quality is the complaint
Usually one of three things: answers that are confidently wrong, answers with no traceable source, or silence where your content actually has the answer.
Handoff loses the thread
The visitor explains the problem twice, once to the assistant and again to a person, which is the moment most of the goodwill is lost.
Migration has no owner
Re-selecting sources and reconfiguring an assistant is a week nobody has scheduled. That is the actual blocker, and it is the part we do.
The teardown
What we test, on your pages
Eight rows, run against both assistants with the same questions. You get the raw answers, not a summary.
| What we check | SiteGPT | Creobot |
|---|---|---|
| Answer correctnessDoes the answer match what the page actually says. | scored | scored |
| Source groundingWhich page the answer came from, and whether it exists. | scored | scored |
| Hallucination riskWhat happens when your content does not cover the question. | scored | scored |
| Citation matchWhether the cited page contains the claim it is cited for. | scored | scored |
| Missing contentQuestions neither assistant can answer, ranked by frequency. | scored | scored |
| Lead intentWhether a buying question is recognised as one. | scored | scored |
| Human handoffWhat a person receives when the assistant steps aside. | scored | scored |
| Widget placementWhere the launcher sits and what it covers. | scored | scored |
HOW THE SCORE IS PRODUCED
Every row is a real answer to a real question, captured as a screenshot with the source it cited. Nothing is inferred from a feature list, and nothing about SiteGPT is asserted beyond what its own assistant returned on your site.
Inputs
What you give us
Your website URL
The public site the assistant answers from.
Where SiteGPT appears
The page carrying the widget, so we test the same surface a visitor sees.
Three to ten questions
The ones your team already knows are asked. Optional; we will pull candidates from your content if you would rather not.
The work
What we do
Teardown of your current assistant
We run your questions against SiteGPT on your live site and record what comes back, including the cases where it declines or invents.
Source inventory and map
Sitemap, docs and help centre, sorted into what should be a source and what should not. This is the step that usually never gets scheduled.
Creobot configured on approved sources
We do the setup. You approve the source list before anything is indexed.
Side by side QA
Same questions, both assistants, captured together so the comparison is yours to check rather than ours to assert.
A recommendation, including no
If Creobot does not produce better grounded answers, cleaner handoff or better lead signal, the recommendation says so and there is no reason to switch.
Risk
How the risk is handled
SiteGPT stays live
Nothing is switched off. The benchmark runs beside your current assistant, not instead of it.
Public content only
Public pages, public docs, public help centre and files you approve. No customer data, no PHI, no payment data, no credentials.
No lock-in during the benchmark
The teardown is not a contract and there is no checkout on this site. Creobot opens to teams in order rather than by open signup.
What you can test without a security review
SCOPE, STATED PLAINLY
Creobot is not SOC 2 certified today. It holds no HIPAA attestation, offers no BAA, and has had no external penetration test. That is exactly why the first benchmark stays public-content-only: public website pages, docs, help content and sources you approve. You can measure answer quality, source grounding and handoff without sending customer data anywhere. Teams whose process requires a SOC 2 report before any pilot at all are a later-stage fit, and we will say so on the call rather than waste your time.
The comparison above is between your current assistant and Creobot on your own content. We make no claim about SiteGPT’s certifications, architecture or roadmap, because we have not audited them and neither should you take our word for it.
Migration questions
Launch
See the gap on your own pages
Send us your site and the page where SiteGPT runs. We test both on your own questions, open every citation, and hand you the transcripts either way. SiteGPT stays live throughout.