The TV Guide used to be the most popular magazine in the United States. My uncle subscribed, and I recall flipping through back issues while he worked the current crossword. The primary purpose of that little weekly book was the listings section, organized by time slot and channel. Networks paid for prime placement, because it was the only way viewers knew what was on. 

Streaming came along and changed the game, bringing us time-shifted, bingeable shows. The “what’s on channel 4 at 8 pm?” appointment television was replaced with “what should I watch next if I liked White Lotus?” 

A similar pattern is emerging now, with developers increasingly using AI over traditional search engines to find new products.

For years, winning your category meant you earned the developer mindshare. The top spot as an analytics database or security monitoring tool gave you a steady supply of developers to nurture into customers. 

That logic still holds for many category queries, but we’ve noticed a shift in LLM recommendations. The old approach breaks down when developers get specific. They’re searching with more context than your category positioning was built to handle.

Category Leaders Don’t Always Survive Context

It’s still a great time to be a category leader. Once a dev tool holds the top spot in developers’ minds, it fades slowly, if at all. The same pattern plays out with LLM recommendations, perhaps even amplified. 

A model’s training has seen so many questions, answers, and odes to top products that they seem to remain in the conversation even after developer sentiment has passed. 

But the more specific queries are less likely to have this same staying power.

LLM Rank tracks hundreds of dev tools in dozens of categories. Prompting models with the real questions developers ask, we’ve seen the leaders hold steady for category queries, but observed a lot of reshuffling when looking for use cases, alternatives, and tech stacks.

In fact, 74% of contextual queries get a new leader, while 30% return a completely distinct top three.

About half of the top 10 are replaced when we compare category to contextual queries across models. In open source, where developers frequently look for products to solve their problems, it’s even clearer. On average, only two category leaders remain in an open source top 10.

In one case, open source searches displaced the entire top 10. They were replaced by:

• Five products from low on the category leaderboard
• Five new products that appeared for the very first time

Every category leader was gone. Only the contextual leaders remained. And contextual queries are growing as developers realize that they can get more precise answers. You want to be on the leaderboard the developer finds when they’re ready to build.

The Query Already Contains the Context

In the SEO era, marketers thought in common search terms. Short, categorical, high-volume keywords were the norm. The job was to translate this minimal information into search intent and eventually convince developers to try your product. While this type of search is not gone, longer queries are steadily replacing it, leaving room for rich context.

One window into this trend is now integrated into Google searches. In about 50% of these, the “top result” is an AI overview. When a user clicks into it, they enter AI Mode, where the response can be further refined. 

Google’s own data shows that these AI queries are getting longer and more specific. They’re asking “what,” “how,” and “can I” questions.

3x longer: Google reports “the average AI Mode search query is triple the length of a traditional search query”

For developers, it means they’re less likely to turn their problem into keywords. They’re providing the full context of that problem within the query itself.

Keywords vs Contextual Search

The search	Words	Characters
auth library	2	12
How do I let users create an account and log back in later with email or their social profiles?	12	95

On Google, the lengthy query is bound to trigger AI Mode. And when coding agents receive these questions, they fire back with recommendations and implementation steps.

Coding agents like Cursor, Copilot, and Windsurf construct queries automatically from a developer’s stack, language, and dependencies. The developers aren’t obliged to translate their problems into a query. It arrives pre-loaded with even more context the developer didn’t have to type.

We’ve tracked similar identity and authentication recommendations from the top LLMs. Consistently, they include these products in the top five: Auth0, Firebase Auth, Okta, Amazon Cognito, and Clerk. Since Okta owns Auth0, there’s a single company frequently fighting with itself at the top of that leaderboard.

But that’s for a category query. Add some additional context and the picture becomes clearer. For example, when developers look for enterprise single sign-on (SSO), they’ll see that Okta now leads the pack. As we saw in the previous section, much of the pack itself has also changed:

Only one other product from the category top 10 makes the SSO-contextual top 10. The rest are mostly filled in by products ranked 11th through 25th. Two new products appear, which were previously invisible across the dozens of category searches we performed.

While it’s useful to see category leaders, the contextual results are where developers envision real solutions to their problems. The context comes from the words the developers use and sometimes the codebase they’re building. The question is whether your positioning shows up inside that context, because that’s what determines whether the LLM recommends your product.

Your Positioning Has Gaps the Data Can Show

Most developer tool companies compete in large categories, such as Cloud Compute or Observability. A handful of category products duke it out for the top few spots, each carving a distinct position. But to what extent does that match what the LLMs think?

Contextual queries show how well positioned your dev tool is with LLMs. AI Coding tools are a clear example of positioning in action, with the gaps visible in the data. 

The category leaders are Cursor, Copilot, and Windsurf. Seemingly missing from that list? Claude Code, which has become a developer favorite. It’s ranked ninth overall: OpenAI only recommends it 60% of the time, Claude ranks itself fourth behind the category leaders, and Gemini doesn’t recommend it at all.

Most of that changes when a contextual query requests an AI coding CLI, the primary interface developers use for Claude Code. Now Claude is the top recommendation: OpenAI suggested its rival in every single one of our tests. Claude gets some confidence and puts itself right up at the top.

Gemini’s silence on Claude Code isn’t an anomaly. This kind of disagreement between models is common in our tests. In fact, it could be a reason that you don’t notice the gaps in your positioning: when you check one model, you only see a single point of view.

50% of the time, asking a second model returns an entirely different top three.

One model is not enough to know if your positioning hits. Most teams don’t perform this sort of sanity check against their positioning. Because this data is publicly available, you can run it on your competitors to see how they perform in their apparent positions, as well as against your product in areas where it should be strong.

Here’s how four AI coding tool marketing teams might react to these results:

• Claude: “We showed up where we’re strongest!”
• GitHub Copilot: “We may slip a little, but we’re still in the CLI conversation.”
• Windsurf: “The IDE context is better for us, anyway.”
• Cursor: “What the heck, why does the Cursor CLI get overlooked?!”

Any one call to an LLM may be nondeterministic, but you can find solid facts within the data if you approach it right. The gaps in your positioning are mappable and often fillable, but only if you look at the complete picture.

How Well Does Your Positioning Survive the Query?

Nobody canceled their TV Guide subscription because they stopped watching television. The time and channel format no longer supported how viewers found a show. Category leadership in dev tools won’t go away, but it’s no longer the whole game. 

When developers look for tools like yours, they’re likely to include context that changes the recommendations they receive. Whether your product appears in those results depends on how well your language reflects how developers describe their problem when they’re ready to build.

EveryDeveloper built LLM Rank to show both category and context in developer search. We work with dev tool companies to help close the gap between how they position themselves and how LLMs actually recommend them. 

Let’s start a conversation about your product and the developers you want to find it.