This survey is missing options, and it should be a multi-select survey for the options it has. Our real answer is, "We use autocomplete frequently in both 'Yes to all' and 'Yes/No' mode, depending on the document."

We often document multiple related products in a product family, and we share bookmaps and/or chapter maps between them. We typically do our primary authoring using the most featured product, and then we autocomplete the other publications against the baseline of the primary publication. We try to keep all the related product manuals on the same version of each topic so that we can use a "Yes to all" completion, but sometimes there are enough inconsistencies between the products themselves that we must select yes/no for each object.

For ISPI documents, we always use autocomplete to roll up to the latest version of all content. Because they contain sensitive, pre-approved legal and certification content, we must always use the latest, most updated version.

For us, autocompleting baselines is a must-have feature, and we would not want to lose any of the existing options. It's critical that we can complete either to the newest or to a specific baseline. We require the ability to select each object in some instances, but the "yes to all" option is a big timesaver when we know that our document strategy allows for it.

Exactly what John Petersen said. There are many variables about when and how to use Autocomplete. This is not a one-sized-fits-all feature. We need full flexibility when using it. 

Related baseline feedback: While you are considering baseline handling for a web-based publication editor, I feel this related feedback applies to the above survey and would definitely affect our use of autocomplete.

Since we started using Tridion Docs in 2011, the lack of an actual check-out/check-in workflow for baselines has led to some problems. When two authors are working in the same publication, or when they are working in two publications that share the same baseline, it is very easy for the two authors to inadvertently overwrite each other's baseline changes. This leads to unpredictable outcomes and a high likelihood of error, where the wrong version of a topic ends up in a publication. There are two behaviors that, when taken together, are problematic:

1. Either author can make a baseline-altering change to the publication and save at any time.
2. The other author's Publication Manager instance neither updates to the newly saved baseline in the repository nor alerts the author that the baseline in the repository has changed.

This means we've largely abandoned the idea of using shared baselines for related publications. We have become dependent on the more predictable behavior of autocompleting baselines, where the primary publication is the "source of truth" for what versions to use, and the other publications are autocompleted intentionally against that.

It seems that a web-based publication editor could have a more real-time awareness of the repository baseline and provide opportunities to keep multiple authors aware of changes and synchronized. If so, this would likely increase our usage of shared baselines and decrease our use of autocomplete.