Your Focus: Fact Checking

Fact checking means no AI fact reaches a buyer until you've opened its source, with the strictest checks on the claims that cost the most. To move up one level, change what triggers your check: from facts that feel surprising to every fact about the prospect, then to checks matched to the stakes. This page shows what each answer looks like, how to tell which one is yours and a 30 day plan to move up one step.

The four levels of fact checking

Each answer word for word, what it looks like in a real week and what keeps people there.

What changes between levels is what triggers the check. At 0, nothing does. At 1, a feeling does: the fact seems surprising. At 2, a category does: every fact about the prospect, however normal it sounds. At 3, the stakes do: you open the exact source behind each prospect fact, and claims about your own company's pricing, security or results go to the person who owns them.

Where you probably are

1. Check your results screen. Every area shows a bar with your score, such as "1 of 3." That number is your level here.
2. Use the tie rule. When scores tie, Fact checking comes first of ten, ahead of every other area. So it's in your three unless three other areas scored lower than it. It may be in your three because it wins ties, not because you scored lowest here. With a perfect score, it's the first of your "next three moves," and the results screen shows its advanced tip instead.
3. Match your profile. Each plan page describes this area at its level.

If Fact checking is one of your three, trust the bar over the profile. If the bar says 2 or 3, it's probably there because it wins ties. Start the plan from your level, not from week 1.

Your move up one level

Aim for one step, not the top.

1. From 0 to 1: pause before you paste

   Target answer: "I double check anything that seems surprising." Before any AI fact goes into a message, read it once and ask: would I bet a deal on this? Check anything that gives you pause. AI predicts likely text, so a confident guess reads like a fact; How AI Assistants Actually Work shows why. This is a first step, not a finish line.

2. From 1 to 2: check by category, not by feeling

   Target answer: "I check prospect facts like titles and recent news against LinkedIn or the company site." The wrong facts that hurt sound ordinary: a title that changed, an office that's planned but not open. So check every fact about the prospect, every time. Names and titles go against the company site or LinkedIn, read yourself. News goes against the company's own newsroom, with the date.

3. From 2 to 3: open the exact source, then sort by stakes

   Target answer: "I match my checks to the stakes: I open every source for prospect facts, and pricing, security or ROI claims get a source plus approval." Two changes. Open the link AI cited, not just LinkedIn: a link isn't proof until it opens the right page. Then add your own company's claims. Find out who owns pricing, security and ROI at your company before you need them.

4. Already at 3: find the weak spots faster

   Your advanced tip: "Ask the model “what in this answer are you least sure about?” before every important send. It is a fast way to find the weak spots." Treat the answer as a map, not proof: you still open what it points to. Build the question into your prompt template so it runs every time. Your AI Plan: AI Power Seller has a prompt that audits five things you sent.

Your 30 day plan

Your results screen gives this area one tip. This page turns it into a month, so checking becomes a reflex, not a reaction. Week 1 starts with your results tip: open every source and keep a log.

Where to check each kind of fact

The first five rows are medium stakes: open the source. The last row is high stakes: source plus approval. It's the row people at 2 usually skip.

Prompts for fact checking

Run these in your company approved AI tool. Each matches a week of the plan or a step up. None of them checks a fact for you: each one tells you where to look.

Here's a draft I plan to send to a {title} at {company}. List every fact in it about a person, an event, a number or a company, including the ones that sound ordinary. For each, give: the fact in a few words, the kind of fact, the source link you used (or NO SOURCE) and the one place I should open to check it.
Put NO SOURCE facts first. Don't rewrite the draft.

<draft>
{paste the draft}
</draft>

Here's my fact checking log for this week: each fact, where it came from, the source I opened and whether it held up.
1. Group the wrong or unconfirmed facts by kind: people, dates, places, numbers, mixed up companies, links or claims about my own company.
2. Tell me which kind went wrong most.
3. For that kind, suggest one line I could add to my prompts or my seller context file to prevent it, and the check I should do first next week.
Use only my log. Don't add examples of your own.

<log>
{paste your log}
</log>

Read this brief on {company} and list only the facts a reader would accept without a second look: job titles, dates, office locations, product names, team sizes and anything described as "recent" or "new." For each, show the exact words, the source you used and the date on that source. Mark it STALE if the source is more than 12 months old, and NO SOURCE if there isn't one.
Don't judge whether they're true. I'll open each one.

Below is my draft and my company's approved materials: pricing, a security summary and approved customer results. List every claim in my draft about our pricing, security, compliance, results or customers. For each, quote my sentence and the approved line that supports it, word for word. If no approved line supports it, write NEEDS APPROVAL and the kind of owner who would sign off, such as finance, security or my manager.
Don't soften or rewrite anything.

<draft>
{paste the draft}
</draft>
<approved_materials>
{paste your approved materials}
</approved_materials>

Act as a {title} at {company} who knows their own company well and reads every email closely. Here's an email I plan to send you. Point to any sentence that would make you think "that's not quite right about us," such as a role, a date, a project or a number, and say why it would stand out. Then name the one sentence I should verify again before I send.

<email>
{paste the email}
</email>

Related focus areas

Each will get its own page; until then, the guide next to it covers the same ground.

• Account research: better sourced briefs mean less to check later. A brief with a date and a link on every point turns checking into clicking. Guide: How to Use AI for Account Research.
• Data safety: checking pulls drafts, CRM records and pricing into AI, so the traffic light applies here too. Guide: AI Policy for Sales Teams.
• Outreach writing: one trigger and one proof point per email means fewer facts to open before you send. Guide: How to Write Cold Emails with AI That Don't Sound Like AI.

Retake the assessment

Retake it at day 30. Answer from your log, not your intentions: if some sends went out with sources unopened, pick the answer that matches what you actually did. One step up here adds one point to your total.

What's next

Everything on this page is free, and so is the guide from your results: How to Use AI for Account Research, whose verification step and log this plan builds on. For why AI gets facts wrong, read How AI Assistants Actually Work and What Is a Hallucination?. To see where checking fits in a whole sales week, read AI for Sales: A Practical Guide. If you want all ten areas, the AI for Sales Certification makes checking by stakes a habit in Session 4, where you correct an AI written brief with planted errors and write your own pre send checklist, then verify three real briefs in Session 5. Lead a team? See team options.

Frequently asked questions

Not sure which reading order fits you? Find your path.