Every message SalesMind AI sends goes through a quality check before it ships. This keeps your outreach professional, clear, and effective.
This article explains what those checks are, why they exist, and what to do when a message does not pass them.
Why quality checks exist
SalesMind AI checks every message for two things:
- Readability: Is the message easy to read? Complex language can confuse prospects.
- Clarity: Does the message sound natural? We remove corporate jargon and sales cliches.
These checks happen in the background. Every message the AI writes goes through them before it sends.
👉 Note: Readability checks (FK score, syllable density, AI-sounding filler) run on English messages only. Universal checks (message length, links, questions, placeholders, and the overall quality review) run on every language.
What is the FK score?
FK stands for Flesch-Kincaid. It is a standard score that measures how hard a text is to read. It maps to school grade levels.
- Grade 1 to 5: Very easy to read (like a children's book)
- Grade 5 to 7: Easy to read (like a news article). This is our target.
- Grade 8 to 10: Fairly hard to read (like a business report)
- Grade 11 or higher: Hard to read (like a legal document)
SalesMind AI aims to keep messages between grade 5 and 7. This is the sweet spot for professional outreach that feels natural and gets replies.
In practice there are two thresholds: above grade 7.5 the message ships with a soft warning; above grade 10 it is rejected and rewritten. These readability scores are only computed for English messages.
Tip: If your messages score above 7, they may sound too formal. If they score below 5, they may sound too casual. The AI adjusts for your Voice DNA, so a formal tone may produce higher FK scores.
What is syllable density?
Syllable density measures how many long words your message uses. A word with 3 or more syllables counts as a complex word.
Messages with high syllable density are harder to read. SalesMind AI flags them so you can simplify your language.
Example:
- "I utilized the aforementioned methodology" has high syllable density (hard to read).
- "I used the method we discussed" has low syllable density (easy to read).
Soft warnings vs hard rejections
When a message does not pass a quality check, one of two things happens:
| Result | What it means | What happens next |
|---|---|---|
| Soft warning | The message is close to the limit but still OK | The message ships with a warning tag. You may see a note in your dashboard. |
| Hard rejection | The message fails the check clearly | The message does not ship. The AI tries again. |
Think of it like a car dashboard:
- A soft warning is like the check engine light. Watch it, but you do not need to stop.
- A hard rejection is like the fuel light. Action is needed right away.
What happens when a message is rejected
When a message gets a hard rejection, SalesMind AI tries up to 3 times to write a better version.
Each attempt:
- Looks at why the first message failed
- Tries to simplify the language
- Checks again
If all 3 attempts fail, SalesMind AI compares every attempt and ships the best one (the version with the fewest and least serious issues). A message with AI-sounding filler never wins over one without it. Only messages that fall to the lowest quality grades are held back from sending. You may see an alert in your Slack or dashboard.
Tip: If your campaign produces fewer messages than expected, check your Voice DNA settings. A very formal Voice DNA can cause the AI to write messages that fail quality checks.
Common rejection reasons
| Reason | What it means | How to fix it |
|---|---|---|
| FK score too high | Message is too hard to read (English only) | Simplify your language. Use shorter sentences and simpler words. |
| Syllable density too high | Too many long words (English only) | Replace complex words with simpler ones. |
| AI-sounding filler | Message uses phrases that sound machine-written, like "caught my attention", "this shows a commitment", "is key", "that kind of", or empty praise like "is impressive" | Rewrite with the concrete fact in plain words. The AI is told exactly which phrase was flagged and avoids it on the next attempt. |
| Forbidden phrase detected | Message uses a banned corporate cliche | Rewrite without the cliche. Check the list of banned phrases in your campaign settings. |
| Ask pattern detected | Message sounds like a hard sell or pitch | Rephrase as a natural conversation starter. |
| No call to action detected | A message ends with no question, link, or invitation | This is a warning only - it never blocks sending. If the step is a deliberate soft close (like a final break-up message), add the marker NO_CTA to that step's instructions and the system will respect it. |
| Call to action too early | An early message pushes a meeting or link before trust is built | Save the ask for a later step. If your step instructions deliberately ask for it, the system honors your instructions. |
| No question in message | A message that should engage the prospect ends without a question | Second touchpoints, connection notes, and steps whose instructions say "no question" are exempt. |
| Low quality grade | The overall quality review graded the message D or F | The message is rewritten automatically. If it cannot be fixed, it is held back. Check that your step instructions do not contradict your campaign settings. |
| Personalization missing | Message does not include the prospect's name or company | The AI auto-personalizes from LinkedIn data. If this triggers, check your Voice DNA and message instructions. |
How Voice DNA affects quality
Your Voice DNA tells the AI how to sound. If your voice is very formal (like a lawyer or executive), the AI will produce messages with higher FK scores. This is expected. It may cause more quality check failures.
Options to improve quality:
-
Edit your Voice DNA: Make it less formal. Remove phrases like "gravitas" or "delivery leadership expert" if they cause issues.
-
Add message instructions: In your campaign settings, there is a Message Instructions field where you can give the AI specific rules to follow for every message in the campaign. This is how you guide the AI's writing style without using static placeholders.
Examples of useful message instructions:
- "Always open by referencing the prospect's company name."
- "Keep every message under 40 words."
- "Use a casual, conversational tone. No corporate language."
- "Each message must end with a single soft question."
- "Never mention pricing. Focus on the problem only."
- "Reference the prospect's industry in the first sentence."
💡 Tip: Message instructions work alongside your campaign objective. The objective sets the goal; message instructions set the writing rules.
-
Use thumbs-down feedback: When you see a message that is too complex or off-tone, give it a thumbs-down. The AI learns from this over time.
When your instructions and campaign settings disagree
The most common cause of repeated rejections is a contradiction between the campaign configuration and the step instructions. Two classic cases:
- The campaign objective is Drive traffic to your website and a Landing Page URL is set, but every step instruction says "no links". The system must place your link in the final call-to-action message, so the quality review keeps failing it.
- The final step is a deliberate soft break-up, but nothing tells the system that - so it keeps flagging the missing call to action.
How to fix contradictions:
- Give the link a home. Exactly one message step should carry your landing or booking link. Say it explicitly in that step's instructions: "End with this exact link: ...".
- Declare deliberate soft closes. If a step must not contain any ask, question, or link, add the marker
NO_CTAon its own line in that step's instructions, for example: "NO_CTA - system rule: this is a final no-reply close by design. Do not include any call-to-action - no meeting ask, no question, no link." This works in any language. - Keep the objective honest. If you do not want links in your messages at all, leave the Landing Page URL empty.
What is next
- Review your Voice DNA if messages are getting rejected
- Train the AI with feedback
- Check why a campaign is not sending
Messages that end with a link
Some of your touchpoints end with a link instead of a question. For example, a final message that invites the prospect to book a call (Calendly), join a webinar (Zoom), or message you on WhatsApp will often end on that link rather than a question mark.
SalesMind AI recognizes this. When a message genuinely ends on or includes a real call-to-action link, it is not rejected for "not ending with a question." The link is the call to action, so adding a question after it would only weaken the message.
What still applies to these messages:
- Readability checks (FK score) still run. A link does not exempt a message from being easy to read.
- Syllable density still applies. Keep the wording simple.
- Length limits still apply (300 characters for connection notes, 400 for messages).
- Quality and safety checks still apply. A low-quality or spammy message is not exempt just because it has a link.
- If your campaign instruction provided a specific link that the message should contain, and the link is missing from the final message, that is still flagged.
In short: a message that correctly ends on its booking, calendar, or contact link will pass the question check, while every other quality rule continues to protect your outreach.