Check out our new The REAL AI Podcast for this week – Episode 4 12/8/25 – AI generated with host Miles and Grace!

REAL AI:  Use AI to verify AI, Prompt perfection, AI stats and facts, headlines, and Quote of the week

By Kevin Hawkins with Korey Hawkins | Vol. 3 Issue 49

A great stocking stuffer: “The REAL AI Guide for Real Estate Agentson Amazon.

Subscribe to our free newsletter here – realai.blog

Use AI to verify AI

AI to verify

Even the best-paid chatbots, including the top versions of ChatGPT, Gemini, and Claude, still make things up. They call it “hallucinating.” I call it what it is: they lie.

That’s because these bots are trained on billions of words scraped from all over the internet. Not every blog post, comment thread, or Reddit rant is accurate. And even when AI is trained with care, the quality of its responses still depends on the quality of the information it learned along the way. Some of it is outdated. Some of it is misinterpreted. Some of it is just plain wrong.

But here’s the thing. Over time, your trust in your chatbot grows. You’ve customized it. You’ve trained it. You’ve taught it your tone, your workflows, perhaps even your client personas.

But let’s not confuse confidence with correctness, especially when your chatbot starts citing statistics or quoting sources you didn’t give it. That’s when the adage applies: “Trust, but verify.”

Brad Inman REAL AI

Challenge your bot to self-check

One of the smartest ways to fact-check AI is with AI.

The first method is simple. Ask your chatbot to double-check its own response and provide an active link to the original source. Emphasis on “original.” Not a recycled blog post. Not a LinkedIn rant. Not a Forbes article quoting a HousingWire stat in a post last month that actually came from a five-year old NAR report.

Real estate is especially prone to stat recycling. A number gets pulled out of context, published without attribution, and suddenly, it shows up everywhere. You’ll see it repeated in newsletters, market updates, and pitch decks.

For example, if you press ChatGPT with a very specific prompt – ask it to review its previous answer, cite each source line by line, and provide live links – you’ll often get a better result. It just takes a little persistence and iteration.

Here’s our REAL AI Rule: If your bot can’t find the original source, or the link it provides is dead, don’t use it.

Your workaround: ask it for a similar fact that can be traced back to the original source with a working link.

Put another Chatbot to work

The second method is to bring in a different chatbot. Yes, fact-check one AI with another.

Because we have monthly paid accounts with the major chatbots, my go-to for real-time fact-checking is Perplexity. Its search results are more current, its citations are usually spot-on, and it handles sources with more transparency.

Google’s Gemini also does a solid job of cross-checking facts. Both typically outperform ChatGPT at verifying specific claims. That’s especially true when you’re checking something ChatGPT itself just told you.

Even the most intelligent bots need a second opinion. And sometimes, that second opinion comes from another bot! (-Kevin)

Prompt perfection

Prompt Perfection

Everyone’s handing out “best prompts” these days: downloadable AI guides, webinar slides, newsletters, and books. Yes, our own The REAL AI Guide for Real Estate Agents is also packed with prompt examples across its 107 pages.

But here’s the thing: what’s best for someone else might be bloated, clunky, or just wrong for what you need. One of the top AI newsletters recently shared its best prompt for generating a professional headshot using Google Gemini’s Nano Banana tech. But the prompt was over 180 words long. That seemed excessive, so I asked Gemini Pro what it thought. Gemini agreed: it was overkill, and it offered a 55-word alternative that produced a better result.

Too many recommended prompts today are what I call “Prompt Salad.” They throw in way more than the chatbot actually needs.

A better solution: ask your paid chatbot to help you perfect your prompt. If you’re using Gemini’s Nano Banana, then ask Gemini to optimize it for you. If you’re using the Swiss Army Knife of chatbots, ChatGPT Plus or Business, use those for your prompt-perfection.

Once you’re dialed in, save that final prompt to your prompt library. And if you use it often, consider adding it to your personalization settings so it’s always ready. (-Kevin)

New AI Facts and Stats

AI Facts & Stats

1. 40% of enterprise software applications will include task-specific AI agents by 2026.

2. 30% of large enterprises surveyed said they require AI fluency training as a condition of employment.

3. 70% of shoppers surveyed stated they are willing to let an AI agent autonomously book flights.

4. 34% of consumers surveyed favor interaction with AI agents when support systems necessitate repeated restatement of information.

5. 93% of business leaders globally agree that scaling AI agents within the next year will be a key competitive advantage.

Source: Secondtalent (-Korey)

New AI Headlines

AI Headlines

OpenAI wants ChatGPT to confess its sins | 12/4/25 Quartz
ChatGPT’s upcoming ‘confessions’ feature will force the chatbot to fill out a second response and explain which rules it broke to answer a prompt.

OpenAI Ordered To Turn Over 20 Million ChatGPT Logs in New York Times Copyright Fight | 12/4/25 eWeek
ChatGPT is losing its privacy and anonymity as the AI copyright law battle heats up.

AI real estate platform Ridley secures $6.4M seed funding | 12/4/25 ColoradoBiz
Ridley is an AI-powered FSBO with DIY tools to sell homes without commission fees.

Google Just Shared How the Next Generation of Leaders Are Really Using AI | 12/4/25 Inc.
Young professionals are embracing AI, but there is a disconnect in trying to personalize AI to fit their style.

This is how poetry broke AI safety | 12/1/25 Mindstream
Writing in poetic structure makes it harder to detect harmful AI prompts.

YouTube’s new AI deepfake tracking tool is alarming experts and creators | 12/2/25 CNBC
YouTube is using biometrics to remove AI-generated videos that exploit one’s likeness. (-Korey)

AI Quote of the Week

Quote of the Week - Michael Tritthart

Free newsletter: Real AI newsletter here – and please share it with a friend!

Content suggestions welcomed: email korey@wavgroup.com