Comparison · Category-level (not a vendor review)
BastionLog vs. asking an AI chatbot
Both answer "what is this company?" They diverge on the form of the answer: a paragraph written from a model's memory, or a dated record assembled from public sources. This page is for comparison only, not a vendor review.
Why this comparison exists
A small trading desk is about to onboard a new supplier in a jurisdiction it has not worked with before. A teammate pastes the company name into a general LLM chat box. The chat returns a confident paragraph with no source URLs, no timestamps, and no indication of which fields came from which database. That paragraph is not safe to attach to a payment file.
General-purpose LLM · When this fits
A chat assistant that writes a paragraph from its training mix. The text reads well; the underlying sources are not shown.
Output is generated on demand from model weights, not from a live query to the named registry. The same prompt re-run later may yield a different paragraph, because the model's sampling or its training mix may have shifted. Useful when the question is "give me a quick orientation", and the reader is you.
BastionLog · When this fits
A structured record assembled from live public sources, each field tagged with provenance. Output is reproducible and dated.
BastionLog queries the named registry and the listed footprint sources directly, records the raw responses, and renders a dated report with source URLs and timestamps attached to each fact. The same query re-run later yields the same factual record. Useful when the paragraph will leave your laptop.
Side-by-side comparison
| Source traceability | General-purpose LLM | BastionLog |
|---|---|---|
| Source traceability | Sources usually not cited. The paragraph reads confident, but the field values are not tied to a URL or a database snapshot. | Every fact is tied to a public source URL with an ISO 8601 timestamp. No un-sourced claim is published. |
| Decision labelling | Often issues an overall recommendation as part of the answer ("looks safe", "I would proceed"). The reader has to decide what evidence the assistant used. | No overall pass / fail, no scoring band, no traffic-light signal. Each source only says whether it was found, not found, or temporarily unavailable. |
| Price & access | Bundled with the assistant subscription (≈ $20 / month) or per-message metered pricing; the marginal cost is small, but the output is not auditable. | Full reports are bought per report (single or in packs). Current prices are on the pricing page. Document extraction is free. |
| Reproducibility | Plain text or chat log. Re-running the same prompt rarely produces the same paragraph, and the result is not reproducible from outside the assistant. | Dated PDF with raw field values, source URLs, and audit-trail entries. Re-running the same query yields the same factual record. |
When each is the right tool
General-purpose LLM · When this fits
Use a general-purpose assistant when you want a quick orientation on a company or industry, when the output is for your own context only, and when you do not need to file the result with a bank or counterparty.
BastionLog · When this fits
Use BastionLog when the output will leave your laptop: as an attachment to a wire transfer instruction, an annex to a supplier onboarding file, or evidence retained for an internal audit. The recipient needs to see the source URL and the timestamp, not a paragraph that may have been written from memory.
Frequently asked questions
Does BastionLog use an LLM at all?
Not on the critical path. BastionLog queries public data sources directly and records the raw response. Any natural-language presentation layer, if used, is confined to optional expression layers and does not alter the underlying facts.
Can I paste a BastionLog report into my AI assistant and ask it to summarise?
Yes. The report is a structured set of facts and source URLs by design, so it is suitable for AI-assisted summarisation, translation, or extraction. We do not run that step for you; your call stays with you.
Will BastionLog catch what an LLM hallucinates?
They answer different questions. BastionLog records what public sources say; it does not adjudicate truth. If a public source says something wrong, the log will carry that fact with the wrong value and the source URL, so you can spot the disagreement yourself.
See also
Ready to leave a paper trail?
New accounts receive 3 full reports once their e-mail address is confirmed (not refilled monthly). Each report covers identity, network, reputation and linked companies; add a BIC or IBAN and it also checks operations. No credit card needed.