AppExchange is often compared to Google and I'll admit, I've said it many times. It is similar in some ways but, vastly different in others.

Salesforce breaks down what's behind their search algorithm into four pillars:

Keywords, Quality & Category

The first pillar is focused on content added to the various fields in a listing. That’s still true. Your title, short description, long description, highlights, and resources all feed relevance.

Historically, Salesforce advised partners to find keywords using the search bar and tools like Google Keyword Planner. However, the AppExchange search bar autocomplete has now been removed, so partners no longer get live keyword suggestions from AppExchange itself 

At the same time, AppExchange now runs on a hybrid search engine powered by Data Cloud, which combines:

• Traditional keyword matching

• Semantic AI search

• Machine learning–based re-ranking 

The challenge is that most partners approach keywords the same way, which leads to very generic descriptions across many listings. When everyone writes the same vague platitudes, nobody stands out, and the algorithm has little useful signal to work with.

It always comes down to messaging, and partners and agencies often overcomplicate this. In my assessments, I tell everyone the same thing: I don’t care what you think you do. I care what you actually sell. Look at current customers and what product or service they paid for. Your real GTM is the one that represents the most logos you have today.

Everyone has a keyword stuffing problem, which will negatively impact you once machine learning evaluates meaning and engagement rather than just word count. One of the first things I do is search for the words “Salesforce” and the company or product name. That alone reveals how noisy or clean a listing really is.

Categories still do not directly impact the algorithm. They exist to help visitors navigate, not to determine ranking.

A major change to be aware of: almost every keyword search now shows “1000+ results” because the new engine retrieves semantically related listings and reports bucketed totals rather than precise counts. You can no longer gauge competitiveness from result counts alone

Engagement

Every listing measures interactions based on number of clicks:

• Screenshots

• Resources

• Videos

• Watch a Demo or Learn More
  
  

The more visitors engage with actual clicks, the more these signals feed into overall search performance.

The highest-valued interactions are CTAs provided by Salesforce. Whether it’s Watch a Demo for ISVs or Learn More for SIs, enable them. SIs should use the native lead capture instead of forwarding visitors to an external website.

When it comes to screenshots, do not put a video as the first screenshot. That’s simply user experience. Extra clicks reduce engagement, many users won’t bother, and autoplay isn’t available anyway.

This is also the point where some people are tempted to “get clever.” Engagement sounds measurable and therefore tempting to automate. Don’t. AppExchange actively detects abnormal, nonhuman activity and spam-style clicking patterns, and attempts to inflate engagement can hurt your listing rather than help it. The algorithm is designed to learn from real buyer behavior, not manufactured signals.

Because the new hybrid engine uses machine learning–based re-ranking, genuine interactions matter more than ever, and synthetic ones are more likely to backfire

Experience

Experience continues to include:

• Partnership level

• Reviews

• Listing and package updates

• Badges such as 1% Pledge or 100% Native

If your listing is missing badges, check the package options in the partner portal or file a support ticket. Respond to every review. Update your package instead of letting it sit unchanged for years.

What changed is how much weight recency carries. Listings that are newer or have been recently updated now receive stronger ranking boosts from the hybrid engine

Learning

This has always been the “secret sauce” part of search, and Salesforce still doesn’t say much about it publicly. That hasn’t changed.

What we now know is clearer: AppExchange search has been rebuilt for the AI era and relies heavily on machine learning trained from user behavior 

Important elements include:

• Recency signals

• Interaction data

• Installs and engagement

• Review volume and patterns

Reviews now act as a stronger ranking lever than before. That brings opportunity, but it also increases the risk created by fake or incentivized reviews already present on the marketplace 

All modern search solutions with ML or AI value strong fundamentals, and AppExchange is no exception. Clear positioning, authentic reviews, frequent updates, and user engagement are not decoration anymore. They are fuel.