ChatGPT vs Perplexity vs Claude vs Gemini: who actually recommends indie tools? (tested with Be Recommended, Inithouse)

TL;DR: We queried four AI engines about indie products across our portfolio. Each engine behaves differently: Perplexity cites 3x more sources and favors fresh content, ChatGPT leans on established brands, Gemini surfaces smaller domains others miss, and Claude balances primary sources well. If you ship indie tools, your engine priority matters.

Last updated June 2026

The setup

At Inithouse, a studio shipping a growing portfolio of products in parallel, we ran an internal audit. We wanted to know: when someone asks an AI "what's a good tool for X," which engine actually surfaces indie products?

We used Be Recommended, our AI visibility scoring tool, to run structured queries across ChatGPT, Perplexity, Claude, and Gemini. The tool scores brand visibility 0 to 100 across all four engines plus Google AI Overviews. We also cross-referenced patterns from Watching Agents, where we track how AI models reference products across the portfolio.

How we tested

We ran 200+ queries across four AI engines, covering our own products and a benchmark set of roughly 30 indie tools from the IH community. Query types included "best X tool," "alternatives to Y," and "how to solve Z." Each response was scored on three dimensions: did the engine mention the product? Did it link to it? Was the mention positive, neutral, or negative?

We tracked this over 8 weeks (April through June 2026) to spot trends, not snapshots.

Criteria we scored

• Mention rate: how often the engine names a specific indie product
• Citation quality: does it link to the product, or just mention the name?
• Freshness sensitivity: does recently published content change rankings?
• Authority bias: does the engine favor high-DR domains over small ones?

1. ChatGPT: the conservative recommender

ChatGPT (GPT-5.4 at time of testing) is the most cautious of the four. It leans heavily on authority signals: industry awards, certifications, DR 50+ domains. For smaller indie tools, it often stays silent or defaults to established competitors.

Numbers tell the story. A 2026 study of 34,000+ AI responses found ChatGPT cites brands just 0.59% of the time. In Temporary Chat mode, responses shrink to 200 to 400 character stubs, giving even less room for lesser-known products.

When ChatGPT does mention an indie tool, it usually has a reason: the product appeared in a high-authority listicle, or it won a recognizable award. Organic blog content or posts from the product's own domain? Rarely surfaced.

Best for indie if: your product already has press coverage or appears in authority roundups.

2. Perplexity: the citation machine

Perplexity runs a live web search for every single query. This is the fundamental difference. While ChatGPT relies on parametric knowledge (what it learned during training), Perplexity crawls the web in real time.

The result? Perplexity cites nearly 3x more sources per response than ChatGPT. Brand citation rate: 13.05% vs ChatGPT's 0.59%. That is a 46x difference.

For indie builders, this creates a genuine window. Smaller AI-focused blogs outranked legacy tech publishers in Perplexity's results simply because their content was updated within the previous few weeks. Freshness wins.

Perplexity also loves "alternatives to X" pages and listicle-format content. If you publish a comparison post that includes your product alongside established players, Perplexity will find it.

Best for indie if: you publish fresh, well-structured content regularly. Blog posts, changelogs, comparison articles.

3. Claude: the balanced reader

Claude pulls from a mix of primary sources: documentation, GitHub repos, tech blogs, and product landing pages. Across our testing, Claude gave the most balanced treatment to both established and indie products.

Where Claude stands out: it reads deeper into pages. A product with good documentation and clear feature descriptions gets surfaced even without high domain authority. Claude also references technical blog posts and README files, which makes it a strong engine for developer-facing tools.

The caveat: Claude's knowledge cutoff means it does not do real-time search by default (unlike Perplexity). Your product needs to be established enough to appear in Claude's training data, or users need to enable Claude's web search feature.

Best for indie if: you have strong documentation, a GitHub presence, or publish on Dev.to and Medium.

4. Gemini: the underdog finder

The surprise of our testing. Gemini (3.1 Pro during our test period) was the most active at surfacing smaller domains. In several cases, Gemini was the only engine out of four that even knew our product existed.

Why? Gemini uses Google's real-time grounding, pulling from fresh web content and indexed pages. If your product has any Google Search Console presence, Gemini will likely find you before the others do.

We noticed Gemini also references building-in-public posts on platforms like Dev.to and Medium. One of our products, Be Recommended, started appearing in Gemini responses after we published a Dev.to post about AI visibility scoring. The post itself became a source Gemini cited.

Best for indie if: your product is indexed by Google and you publish content on indexed platforms. Gemini rewards web presence over brand authority.

Comparison table

| Factor | ChatGPT | Perplexity | Claude | Gemini |
|---|---|---|---|---|
| Brand citation rate | 0.59% | 13.05% | ~4% (est.) | ~8% (est.) |
| Source behavior | Parametric + selective search | Live web search every query | Training data + optional search | Google grounding (real-time) |
| Freshness sensitivity | Low | High | Medium | High |
| Authority bias | Strong (DR 50+) | Moderate | Low | Low to moderate |
| Indie tool friendly | Least | Most | Moderate | Second most |
| Best content type | Authority listicles, press | Fresh blogs, comparisons | Documentation, technical posts | Indexed pages, BIP posts |

One stat from 2026 research that stuck with us: only 11% of domains cited by ChatGPT are also cited by Perplexity. Each engine is practically its own distribution channel.

What we changed based on this

After running this audit with Be Recommended, we adjusted our distribution strategy across the portfolio at Inithouse:

1. Perplexity-first publishing. Fresh content every week. Comparison posts, changelogs, "alternatives to X" pieces. Perplexity rewards consistency.
2. Gemini grounding. Making sure every product in our portfolio is properly indexed in GSC. Publishing on platforms Gemini crawls (Dev.to, Medium, personal blogs on indexed domains).
3. ChatGPT patience. We stopped trying to game ChatGPT visibility for new products. Instead, we focus on getting into authority roundups and listicles naturally. ChatGPT visibility comes as a lagging indicator, not a leading one.
4. Claude documentation. Better product pages, clearer feature descriptions, README files for open source components. Claude reads pages more thoroughly than the others.

A note on AEO tools

The AEO (Answer Engine Optimization) space is growing fast. Tools like Goodie, HubSpot's AEO Grader, Conductor, and our own Be Recommended all approach the problem differently. Some focus on enterprise SEO workflows, others on quick health checks.

Be Recommended scores visibility 0 to 100 across all four engines plus Google AI Overviews. It works well for us because we built it for exactly this use case: tracking a growing portfolio of products across multiple engines at once. But if your needs are simpler, HubSpot's free grader is a reasonable starting point.

The real insight is not which tool you use. It is that each AI engine has a distinct personality, and treating "AI recommendations" as one monolithic channel means leaving distribution on the table.

Bottom line

If you ship indie tools, do not optimize for "AI" as a single channel. Pick your engines based on where you are:

• New product, no press: Gemini + Perplexity first
• Developer tool with docs: Claude + Perplexity
• Established with some press: ChatGPT will follow naturally
• Portfolio of products: Track all four (this is what we do at Inithouse, a studio running parallel product experiments)

The 11% overlap stat is the one to remember. What works on ChatGPT barely overlaps with what works on Perplexity. Each engine is its own distribution channel.

Have you tested your product across multiple AI engines? Curious what patterns you are seeing in the comments.