In the lesson you run Cedar & Pine Marketing, a made-up agency pitching a regional credit union. Deep Research writes a plan, reads dozens of websites for about five minutes, and hands back a long cited report. It looks thorough. The mistakes below are about the gap between looking thorough and being right.
Mistake 1: Why does a broad request waste a Deep Research run?
"Tell me about credit unions" gets you a long report about everything. Fix: say who the meeting is with, what you need to find out, ask for citations and keep it short.
The lesson's prompt does all four: a pitch meeting next week with a regional credit union; how US credit unions use AI chatbots and email marketing today; the 3 most common use cases, typical results, compliance and privacy concerns, and 5 smart questions to ask; "Cite every claim. Keep the report short."
Mistake 2: Why read the plan before the research starts?
Because it's your only chance to fix the search before Gemini spends minutes on it. Fix: click More to read every step, then Start research or Edit plan.
The plan shows only three steps until you click More. In the lesson, step 4 covers NCUA rules, GLBA and TCPA, which is exactly what a credit union cares about, so the plan is right. If it weren't, Edit plan lets you say what to change in your own words. And if you do nothing, it starts by itself after a few seconds, so don't look away.
Mistake 3: Why are very exact numbers a warning sign?
Exact numbers feel trustworthy, but they can reveal a tiny sample. Fix: look at the numbers before you believe them.
The lesson's report opens with 41.67%, 8.33% and 58.33%. They're all multiples of 8.33, which is what you'd get from about twelve answers. Nothing in the report says so. Spotting it takes ten seconds of looking, and it changes whether you'd say the number in front of a client.
Mistake 4: Why isn't a citation the same as a fact?
A citation only tells you where to look. Fix: click the small arrow after the paragraph, open the source card and read the source.
In the lesson, the source behind those numbers is a blog post on a vendor's website. It credits the figures to a "CULytics survey of credit union leaders" but, in the lesson's test, never says how many leaders answered. The citation was real. The number still wasn't safe to quote. The source verification glossary page covers the habit, and AI citations explains what citations prove.
Mistake 5: Why check who published the source?
A vendor's blog post about its own market isn't an independent study. Fix: note who published each number you plan to use, and say it out loud.
If the only source for a big number is a company that sells into that market, you can still mention it, but as "one vendor's survey," not as an industry fact. The lesson's advice: name the source out loud in the meeting. A client who hears where a number came from trusts you more, not less.
Try it free, step by step
Gemini Deep Research: Build a Cited Brief Before a Client Meeting. Ask Gemini Deep Research to research a prospect's industry, read its plan before it starts, then check where the key numbers really came from. You do every step yourself in a practice copy of Gemini, it checks your work as you go, and “Do it for me” finishes any step you get stuck on. Free, and nothing touches your real accounts.
Start the free lessonMistake 6: Why not bring the whole report to the meeting?
Nobody reads a long report in a pitch, and every extra claim is one more you haven't checked. Fix: export it to Docs and cut it to one page.
Under Share & Export, Export to Docs opens a new Google Doc in a new tab, named after the report. Edit it down to the points you'll use and keep only the numbers you checked. To rehearse, Create can turn the same report into flashcards or an Audio Overview.
Mistake 7: Why be careful with Share?
Share makes a link, and a link can travel. Fix: use Export to Docs or Copy for your own work, and Share only when you mean to.
The lesson points this out at the export step. A research report about a prospect's industry is usually for you, not for them, at least until you've checked it.
How do you avoid these mistakes as a team?
If several people prepare for client meetings with Deep Research, agree on a short routine:
- Every request names the meeting, the questions, "cite every claim" and "keep it short."
- Someone reads the full plan before it starts.
- Every number used in a meeting has been traced to its source, with the publisher and sample size noted.
- Vendor-published numbers are labeled as such.
- The brief that goes into the meeting is one page.
The lesson notes that Deep Research needs a Gemini plan that includes it and that Google changes limits often, so check Gemini's own help page before you rely on it. The lesson is part of the free AI for Sales and AI for Meetings courses. For requests to start from, see Gemini Deep Research Prompts.
Key takeaways
- Name the meeting, the questions, ask for citations and keep the report short.
- Read the whole plan with More, and fix it with Edit plan before it starts.
- Very exact percentages can mean a very small survey.
- A citation is where to look, not proof: open the source and check who published it and how many people answered.
- Cut the report to one page and quote only what you checked.
Frequently asked questions
- Can Gemini Deep Research be wrong?
Yes. Its report is only as good as its sources, and a source can be a small or one-sided survey. Check the numbers you plan to use.
- How do I see Gemini Deep Research's sources?
Click the small arrow after a paragraph to see its citations, then click a source card to open the page in a new tab.
- What does Edit plan do in Gemini Deep Research?
Gemini shows the plan again and asks what to change, and you answer in your own words before it starts searching.
- Is Cedar & Pine Marketing a real agency?
No. It's made up for the lesson, and the report text is shortened from a real run.