You can see a word on a scanned newspaper page, yet a search for it returns nothing. The search may be using recognised text that differs from the image. That gap matters when you are tracing an advertisement or a product name.
Keep image and recognised text apart
The Library of Congress technical guidance explains how optical character recognition supports newspaper text searches and how recognition errors affect results.
A search result is therefore a useful lead, not a guarantee that every printed occurrence has been found. Faded type, unusual layout and other features can complicate the text layer.
Try a controlled search note
Record the newspaper, date range, spelling and filters you used. Try a documented alternative name or a shorter distinctive phrase, changing one condition at a time.
When you find a possible match, open the page image and verify the wording and context. A snippet may join fragments or omit the neighbouring explanation that changes its meaning.
Describe the limit honestly
Write “not found using these searches” rather than “never advertised”. Preserve the issue date and page reference for a verified result so someone else can locate it.
Once an advertisement is located, read its claims as claims. Better retrieval solves the problem of finding evidence; it does not automatically establish that the advertiser’s promises were true.