How you record any search (whether productive or not) depends upon your purpose in carrying out the search.
Testing a proposition
If you had an assertion (or testable proposition) in mind such as A and B were married at X in the period 1800-1840 then
the results of your search, whether new data or empty set, need to be attached to (or at least linked to) that assertion.
Using Gramps, I will have a (putative) family entity for A and B with a (presumed) marriage event on which I would
create a note of type Research. There I would record where, when, and why I searched with a comment on possible reasons for the lack of success and implications for future action.
If you wish you can create a note type of your own called "Empty Search". At some future time you can filter all your notes for a sorted list of empty searches.
Exploring a resource
There are times when a family historian will carry out "research" without a particular assertion in mind. In my case, this occurs when a favourite repository announces
the availability of new datasets. If Trove at the National Library of Australia releases a new digitised regional newspaper, I will explore it by selecting some ancestral names that could have been associated with the location and running a few searches. This is just beachcombing, or butterfly catching. If I see something pretty, I will take it home but the only conclusion I can draw from coming home empty-handed is that (on that day) the source was not productive.
In those circumstances, any record I make is a precaution against being drawn back by whatever attracted me in the first place. Next time I decide on a wander through old newspapers to clear my mind, I need to be able to avoid paths I have followed before (or to retrace them deliberately, if there has been a suitable time lapse).
In Gramps, I have a choice of either Repository object (for Trove) or Source object (for the Bullamakanka Weekly Bulletin) as the place for the note that outlines my exploration (particularly the date) and its lack of productivity. In the case described, my personal practice is to use the Repository
to hold that, since I won't create the Source record until it produces something I use.
Neither of the above discussions has any particular bearing on negative evidence, but when you review all of the research notes you have made on an asserted event, you may see a pattern of unproductive searches emerging
that, taken together, challenge one or more aspects of your assertion. Perhaps A and B were common law partners, or migrated to X from Y, or married very young, or ... . A single failed search will not make you modify the assertion but
continuing failure to find support will cause doubt to grow.
Of course the ultimate piece of negative evidence comes from a search which is unsuccessful but not because it is empty. If your search locates a record that A married C at X in 1822 then your proposition is seriously challenged.
Unless you have other information that now becomes conflicting evidence, and perhaps there was a second marriage ...