AI competitor price change research: save the evidence before summarizing
AI competitor price change research cannot establish a change from a current pricing page alone. Save each competitor’s public page, record its billing conditions, and compare it with a dated earlier original. Then check every AI summary against those records. The useful deliverable is a source map that separates supported changes, unresolved questions, and interpretations.
Capture: Preserve the pricing page and the conditions attached to its offer.
Compare: Check equivalent plans against dated earlier evidence.
Classify: Mark each conclusion as a verified change, unconfirmed, or inference.
Verified run
Command: curl -sS -o /dev/null -w status=%{http_code} bytes=%{size_download} time=%{time_total}s -m 20 https://builderlog.net/website-launch/
Environment: Darwin 25.5.0 / python 3.12.13
Ran at: 2026-09-13T17:54:07+09:00
Exit code: 0
status=200 bytes=23093 time=1.118320sThe block below is the captured output of that run, pasted unchanged. A different environment can produce a different result.
The evidence starts with an empty comparison
This is a procedure for a field test, not a report of a completed experiment. The supplied material contains no competitor pricing captures, historical offers, or checked AI summaries. Claiming that a price changed would manufacture the very receipt this method is designed to preserve.
| Evidence item | Available record | Conditions and scope |
|---|---|---|
| Preparation date | 2026-09-13 | Document generation date; not a competitor observation date |
| Reviewed or tested date | Not established | No completed pricing review was supplied |
| Current pricing evidence | Not supplied | No original page capture to inspect |
| Earlier pricing evidence | Not supplied | No supported before-and-after comparison |
| Research scope | Proposed manual comparison | Public pricing pages and their attached terms |
| Supported finding | Evidence is insufficient | No price change or recent trend can be reported |
The table is an evidence inventory. Its blanks matter: they prevent a preparation date from quietly becoming a tested date.
Use the workflow below to fill those gaps. Until then, the defensible conclusion is not established, rather than “prices stayed the same.”
Save the offer, including the small print
Open each selected competitor’s public pricing page in your browser. Use neutral labels such as Competitor Alder, Competitor Birch, and Competitor Cedar in the working document. These are placeholders, not observed businesses.
Before saving, record the page address and your observation date. Note the selected currency, billing interval, region if shown, and whether the page presents a public offer or requires an account.
Use the browser’s print function to save a PDF. Check the saved file immediately. If it omits a pricing panel or footnote, preserve a screenshot of that section as well.
Include the plan name, displayed amount, billing wording, included allowance, and relevant restrictions. A cropped amount without its conditions is a weak receipt.
If a billing selector changes the offer, save the relevant states separately and label them. Keep the displayed wording intact. A monthly equivalent attached to annual billing should not become a monthly payment in your notes.
Suggested artifact caption: “Original pricing-page capture, with observation date, selected billing interval, currency, and the footnote governing the displayed offer.”
The caption belongs beneath a real capture once collected. It is not a substitute for one.
Find the earlier offer before declaring a change
A current page answers what is displayed under the recorded conditions. Establishing a change requires an earlier comparable record.
Look for an existing saved copy of the original page or a dated announcement from the competitor describing the previous and replacement terms. Keep the original address with every record.
Separate observation date from effective date. The day you save a page is not necessarily the day its pricing took effect. An announcement’s publication date may also differ from its implementation date.
Compare the underlying offer, not merely the plan label. Check billing interval, currency, included usage, minimum commitment, and customer eligibility. If those conditions differ, explain the mismatch before calculating or describing a movement.
When no earlier original is available, record the current offer and mark the change unconfirmed. Do not reconstruct a previous price from an unattributed summary.
A newly saved page proves an observation, not a newly introduced price.
Calling something a recent trend needs dated original evidence within the research brief’s recency window. A fresh capture alone does not meet that requirement. Even a supported change at an individual competitor does not automatically establish a broader market direction.
Let AI organize the records, then inspect every claim
Once the captures are readable, use AI to organize their contents into a comparison. Limit the source material to the records you have collected.
The useful output is a structured extraction: competitor label, plan, displayed offer, billing basis, included allowance, eligibility, source location, and unresolved fields. Leave missing information visibly unresolved.
Keep extraction separate from interpretation. First establish what the source says. Then assess whether the earlier and current records support a change.
Inspect each extracted statement beside the original. Check amounts and currencies character by character. Read qualifying footnotes. Confirm that the cited section supports the conclusion rather than merely mentioning the plan.
If a saved page is unreadable or incomplete, return to the source and repair the capture. A fluent summary cannot repair missing evidence.
Avoid uploading account details or unrelated personal material. This procedure needs public offers and their terms.
The summary is a navigation aid; the original page carries the evidence.
Build a source map someone else can check
A source map connects each conclusion to its supporting records. A plain document or spreadsheet is sufficient; no coding is required.
Copy this structure for each claim:
| Field | What to enter |
|---|---|
| Competitor and offer | Neutral label and exact plan wording |
| Earlier record | Original address, saved filename, observation date |
| Current record | Original address, saved filename, observation date |
| Relevant passage | Exact section, excerpt, or screenshot location |
| Comparison conditions | Currency, billing interval, allowance, eligibility |
| Effective date | Source-supported date, or unknown |
| Classification | Verified change, unconfirmed, or inference |
| Conclusion | Narrow statement supported by the records |
| Open question | Missing evidence needed to resolve uncertainty |
Use verified change only when the originals establish a difference under comparable conditions. Use unconfirmed when evidence is missing, contradictory, or not comparable. Use inference for an interpretation that extends beyond the observed difference.
For example, an apparent shift toward larger customers would be an inference unless a source explicitly established that intention. Keep it separate from any observed packaging change.
Copyable review checklist:
- Preserve the current page and its original address.
- Record observation date and visible offer conditions.
- Locate a dated earlier original.
- Check that the offers are comparable.
- Match every AI extraction to a source passage.
- Separate observation dates from effective dates.
- Assign a classification to each claim.
- Leave unsupported fields unresolved.
- Attach readable captures before sharing conclusions.
The final decision follows the weakest missing receipt
No execution failures were supplied for this article. The following are review risks, not reported incidents: a printout can omit content, a selected billing state can disappear from a capture, and a summary can drop a qualifying condition.
A public offer may also leave negotiated pricing or existing-customer treatment unresolved. Record those boundaries instead of extending the comparison beyond its scope.
Final decision: Use manual captures and a checked source map for this research task. Report a price change only when comparable original records support it. For the material supplied here, no competitor change or recent pricing trend is verified.
Related build logs
Save comparable originals, check the AI extraction, and let the source map distinguish verified change, unconfirmed, and inference.