Article Summary
When AutoRFP generates a response, it doesn't just search your content for similar text — it also looks at tags. Every project you create carries tags (like Product: Platform X or Region: EMEA), and so does your content. For each question, the response engine compares the two and uses your settings to decide what a tag mismatch means: should mismatched content be excluded entirely, or just ranked lower?
In the Response Engine settings is where you make that choice, per tag category. It's also where you give the engine general context about your organization, set a quality bar for generated answers, and decide how new responses get tagged so your library stays organized on its own.
Prerequisites
An AutoRFP admin account
At least one tag category with tags, created under Content → Categories & Tags
Content in your library that's tagged — tag rules can only work if your content actually carries tags
Step-by-Step Instructions
Step 1 — Open the Response Engine settings
In the left sidebar, scroll to the bottom section and click Organization (the building icon). This opens your Organization Settings.
Across the top of the Organization Settings page you'll see a row of tabs. Click the Response Engine tab.
Step 2 — Set your quality bar
The Requirement Satisfaction Threshold (0–100) is a quality gate. After drafting a response, the engine scores how well it actually satisfies the requirement — if the score falls below your threshold, the response isn't used.
Default: 20
0 disables the check — every generated response is used
Raise it if you'd rather leave a question blank than accept a mediocre auto-answer; lower it if you want maximum coverage and plan to review everything anyway
Step 3 — Decide on Project Refinement
In Post-Processing, the Project Refinement toggle adds an extra review pass that re-reads and polishes each generated response before presenting it. Better polish, slightly longer generation time.
Step 4 — Rank your content sources
In Content Priority, set each content type to Low, Normal, or High:
📚 Content type | 🔎 Default |
Library | High |
Past Projects | Normal |
Documentation | Low |
Web Scraped content | Low |
Priority is a ranking boost, not a filter — a Low-priority document can still be used if it's clearly the best match, but when quality is comparable, the higher-priority source wins. A good rule: your curated Library stays High (it's your approved voice), web-scraped content stays Low (useful backup, not the voice of record).
Step 5 — Choose a tag strategy for each category
Each tag category gets its own card with two option cards — pick one per category:
1) Filter by tags — "Only use tagged content."
Content that doesn't match the project's tags in this category is excluded from consideration entirely, no matter how relevant its text is. Content with no tags in this category is excluded too. Choosing this reveals a Filter mode dropdown:
At least one matching tag (default) — content is kept if at least one of its tags in this category lines up with the project's selection, even if it also carries other tags
Only matching tags — the strictest setting: content is kept only if every tag it carries in this category matches the project's selection (or is a broader parent of it)
2) Prefer tag matches — "Prioritize matching content."
Nothing is excluded; instead, tag match quality adjusts ranking. Fully matching content keeps its full score, and content is progressively penalized the worse it matches (partial match → untagged → related-but-different → unrelated). Choosing this reveals an Importance dropdown:
High — mismatches are penalized heavily; this category dominates ranking and takes extra weight from lower-importance categories
Medium (default) — a balanced influence alongside text relevance
Low — a mild nudge; the engine stays free to use whatever answers best
Each category is independent — you can be strict about Product and relaxed about Industry at the same time.
Step 6 — Choose how new responses get tagged
The engine's answers become reusable content, so they need tags too. Per category, the "Assign tags to new responses" dropdown controls this:
Using AI Auto-tagging — the AI reads the question and answer and picks fitting tags, but only from the tags your project selected in this category, so answers never wander outside project scope
From the content used (matching project tags) (default) — the response inherits tags from the content items actually used to build it, cross-checked against the project's tags; if the sources contributed no tags, it falls back to the project's own tags
Manually — no automatic tagging; only tags your team applies manually
💡 Tips & Best Practices
Tag your content before tightening your settings. Any strategy is only as good as your tagging coverage.
Start every category on Prefer + Medium. You get tag-aware ranking with zero risk of starving the engine. Promote a category to Filter only where mixing content would be wrong, not merely suboptimal.
Leave auto-tagging on the default unless you have a manual curation workflow — it keeps your library organized without extra effort.
✋🏼 Common Mistakes to Avoid
Enabling strict filtering with thin tag coverage. If your project selects Product A and only five items are tagged Product A, the engine has exactly five items to work with — everything else is invisible. Thin or missing answers after enabling Filter mode usually mean a tagging gap, not an engine problem.
Setting every category to High importance. If everything is High, nothing is. Extra weight for High categories comes at the expense of the others — reserve it for the one or two categories that genuinely matter most.
Expecting Content Priority to exclude sources. Low priority holds content back in ranking but doesn't remove it. If a source must never be used for certain projects, that's a job for tags with Filter mode.
Need Help?
💬 Live Chat: Available in-app
📚 Learning Centre: learn.autorfp.ai/en