Stop the Scroll: My AI Workflow for Figuring Out What’s Working on Social Media

Social media is one of the hardest marketing channels for restaurants. You can’t live without it, and yet you’ve got to do it every day. You have to stay creative. You need calls to action. It has to matter. It takes so much work that it can’t just be something you do on the side. It has to create value, and it has to be done the right way.

Unfortunately, the algorithm does what it wants. Consumers’ tastes, preferences, and behaviors change every day, and mastering social media consistently is incredibly difficult.

That’s why you need some kind of system to guide you when you create content. You wouldn’t build a LEGO set without the instructions, and you shouldn’t build your social media strategy without understanding what is working, what is not, and what your audience actually responds to.

I’ve been creating daily content on social media for decades. I’m also part of an advanced creators lab focused on best practices, performance analysis, and understanding the clues social platforms give us about what is working and what to do next. I’m living in a state of constant understanding and improvement.

Using what I’ve learned from both, I’ve created an AI-powered instruction manual to help you understand what’s working in your social media content, uncover what’s causing people to stop scrolling, and create more of the content your guests actually want to see.

There are two parts to this process.

The first is analyzing your content performance to understand what is already working.

The second is improving the on-screen titles and opening hooks that convince people to stop scrolling long enough to see what you created.

Let’s start with the data.

Too many restaurants post something because the social media manager likes it, someone on the team thought it was a good idea, or they felt like they needed to post something that day.

That is not a strategy.

If you want to know what your guests, followers, and potential customers respond to, you have to analyze your content performance.

You can prepare that information in one of two ways. You can gather it manually using the analytics available inside each social platform, or you can export it from a social media reporting tool.

The finished dataset should be essentially the same either way.

If you are only using the native analytics inside Instagram, Facebook, or TikTok, use this process:

You can also take screenshots of the content insights shown inside the native app and upload those screenshots into an AI tool like ChatGPT. Ask it to extract the visible information and organize it into a table or spreadsheet.

This can save time, but it is not completely foolproof. Screenshots can create problems when information is cut off, labels are missing, numbers are abbreviated, or several posts look similar. Review the finished table against your screenshots before relying on the analysis.

For the cleanest analysis, a spreadsheet with one clearly labeled row per post is still the ideal format.

Depending on your platform and subscription, tools such as Metricool, Hootsuite, Later, Sprout Social, and other social media management platforms may allow you to export post-level performance data as a CSV or Excel file.

The exact menus, metrics, and export formats will depend on the platform and plan you use, but the basic process is:

Whether you build the spreadsheet manually or export it from a reporting platform, your goal is the same. You want one clean dataset that represents every piece of content published during the 30-day period.

Once the information is organized, upload the CSV or Excel file into a new conversation with ChatGPT or another AI tool that can analyze spreadsheets.

Then use this prompt:

Once you have the results, you should have a much better understanding of what your audience responds to, which formats deserve more attention, and what you may need to stop doing.

The next step is to take what you learned and do more of what is working and less of what is not.

Unfortunately, this process still cannot fully explain the most important part of social media performance.

Your performance data can tell you which posts earned the most reach, watch time, shares, saves, or comments. It cannot automatically explain exactly why somebody decided to stop scrolling and pay attention to one post instead of another.

Every day, your guests scroll past hundreds, and possibly thousands, of videos, photos, and carousels. If they never stop to look at yours, it does not matter how incredible the food looks, how useful the message is, how funny the video becomes, or how much work went into creating it. They never gave it a chance.

Two of the most important creative elements responsible for earning that initial attention are the on-screen title and the opening hook.

The on-screen title is the first prominent text someone sees when your content appears in their feed.

The opening hook is what they hear during the first few seconds, whether it is delivered directly to the camera, spoken through a voiceover, or communicated through another audio element.

Neither should be treated as an afterthought.

Your title and hook need to communicate why somebody should pay attention before they have invested any attention in the content. They can create curiosity, make a promise, introduce a problem, challenge an assumption, target a specific audience, show something unexpected, or make the value of continuing immediately clear.

The right structure will depend on the story you are telling, but you do not need to start from a blank page every time.

These are formulas, not finished titles. You still need to add the specific detail that makes the content worth watching.

[TYPE OF GUEST], YOU’RE DOING [SOMETHING] WRONG

I DIDN’T EXPECT TO FIND [DISCOVERY] AT [PLACE OR OCCASION]

THIS MIGHT BE THE BEST [ITEM, EXPERIENCE, OR IDEA] FOR [SPECIFIC PURPOSE]

HERE’S WHAT HAPPENED WHEN WE [ACTION OR EXPERIMENT]

EVERYONE THINKS [COMMON BELIEF], BUT [CONTRADICTING TRUTH]

WHAT REGULARS KNOW ABOUT [RESTAURANT, MENU, OR EXPERIENCE]

[NUMBER] [ITEMS, STEPS, OR LESSONS] THAT CHANGED [RESULT]

WHAT TO ORDER WHEN YOU WANT [CRAVING, OCCASION, OR OUTCOME]

WHAT IT TAKES TO [DELIVER A RESULT OR CREATE AN EXPERIENCE]

WOULD YOU CHOOSE [OPTION A] OR [OPTION B]?

A restaurant could use these formulas to promote a menu item, explain a process, recruit employees, announce an offer, tell a founder story, highlight a guest experience, show community involvement, introduce a new location, explain a loyalty benefit, or build anticipation for an event.

The formula creates the structure. The specific story makes it interesting.

The same principle applies to what someone hears first. These are opening structures that should be completed with real information from the video.

“If you’re the kind of person who [SPECIFIC BEHAVIOR OR PREFERENCE], you need to see this.”

“Before you [VISIT, ORDER, OR MAKE A DECISION], you need to know this.”

“I did not expect to find [SPECIFIC DISCOVERY] here.”

“Everyone tells you to [COMMON ADVICE], but I think that is the wrong move.”

“In the next few seconds, I’m going to show you how to [DESIRED OUTCOME].”

“Would you try [UNUSUAL ITEM, EXPERIENCE, OR COMBINATION]?”

“Here’s what the regulars know about [ITEM, OFFER, OR EXPERIENCE].”

“If you struggle with [PROBLEM], this is what you should do instead.”

“This might be the best [ITEM, EXPERIENCE, OR SOLUTION] for [SPECIFIC PURPOSE].”

“We tried [ACTION OR EXPERIMENT], and this is what happened.”

The point is not to force every piece of content into one of these exact sentences. The point is to stop treating the beginning of the video as an introduction.

The opening is not where you clear your throat, introduce yourself, or slowly explain what the video is about. It is where you give somebody a reason to stay.

Your AI tool, whether that is ChatGPT or another LLM, should understand your brand, audience, voice, goals, offers, and the type of language you would and would not naturally use. Otherwise, it will probably give you generic ideas that could belong to any restaurant.

Once that foundation is in place, upload this article into a new chat along with the finished script, content concept, or creative brief you are working from.

Then use this prompt:

The goal is not to ask AI to replace your creativity. It is to give yourself a creative thought partner that can help you explore more ideas, apply different angles, and get beyond the first obvious answer.

You still make the final decision.

You cannot improve your social media by blindly creating more of it.

You need to understand what is performing well, look for the patterns behind that performance, and use those lessons to make better decisions about what to create next. Then you need to make sure people actually stop long enough to see it.

That means analyzing your performance, improving your on-screen titles, strengthening your opening hooks, and continuing to test what earns attention from your specific audience.

Have you tried a process like this before? Do you have a better method?

If you need help with how to use AI or anything else restaurant marketing related, send me an email at [email protected]

- Rev Ciancio

WHAT DOES REV DO?