I've been conducting a series of workflow efficiency benchmarks for various AI writing assistants, and a recurring point of friction has surfaced with Grammarly. When documenting technical specifications or company literature, it aggressively flags our registered trademarks (`MyCorpAI`, `DeepLens`, etc.) as spelling errors and suggests incorrect "corrections."
This is a known issue in automated editing tools, but it significantly impacts the time-to-completion metric in my tests. Manually dismissing each suggestion for a document filled with product names is not a scalable solution.
I am aware of the "Add to dictionary" function for standard words, but this doesn't seem to apply correctly to multi-word proper nouns or trademarked terms. My current workaround is to disable Grammarly entirely for certain documents, which defeats the purpose.
Has anyone in the community developed a reliable method to address this? I'm looking for a systematic solution, such as:
* A dedicated "Ignore" list for phrases within the Grammarly application settings.
* A method to define custom rules that override spelling checks for specific capitalized terms.
* Confirmation if the Grammarly Business tier offers better terminology management controls.
My ideal outcome is a configured environment where I can run Grammarly on technical documents without it introducing noise by challenging validated proprietary terminology. Any insights into configuration files, undocumented flags, or proven workflows would be valuable for my ongoing benchmark analysis.
BenchMark
Yeah, the add-to-dictionary trick is hit or miss with trademarks. For a systematic fix, you'll want the Grammarly Business account.
It has a Style Guide feature where you can add custom rules. You can list your trademarks (`MyCorpAI`, `DeepLens`) as "Allowed words," and they won't be flagged for spelling. It also lets you set a rule to ignore text inside specific formatting, like brackets, which could help for technical docs.
The downside is it's a team/admin setting, not a personal one. If you're flying solo on a personal account, there isn't a good built-in list. My workaround before we got Business was to create a text file of our trademarks and run a quick find/replace check after the Grammarly pass.
Automate the boring stuff.
I've run similar benchmarks on AI writing assistants and the trademark noise is a measurable drag on throughput. The Business Style Guide approach works, but it introduces a dependency on admin privileges and doesn't help if you're writing for a client or a third-party review process where Grammarly is running on their side.
One thing I've tested that avoids the "add to dictionary" flakiness with multi-word terms is to wrap the trademarked text in a zero-width non-joiner or a non-breaking space between the words, then configure Grammarly to ignore formatting anomalies. That's hacky and fragile across versions.
A more reliable systematic method I've used: pre-process the document with a simple regex to insert a Grammarly ignore comment (if your editor supports it) or to replace the trademark with a placeholder that's already in the dictionary, then swap back after the Grammarly pass. For example, replace `MyCorpAI` with `MyCorpAI` but with a hidden Unicode character that Grammarly skips. The downside is that spell-check-override rules in Grammarly's API don't exist for personal accounts, so you're stuck with either a build step or a post-processing script.
Have you tested the latency impact of a find-and-replace pipeline versus the Business account approach? I'd be curious to see the numbers.