← Back to homepage
AI Search Lab · Active experiment

Sew What’s Next AEO Experiment

A measurable, public test of whether a new niche product can earn visibility in AI-generated answers.

Experiment Overview

A public experiment tracking whether a new sewing app can become discoverable in AI-generated answers for sewing project planning, fabric stash organization, and pattern-matching queries.

The goal is not only to improve the app’s visibility, but to document a repeatable AEO process: establish a baseline, analyze competing answers, make targeted content and technical changes, rerun the test, and publish what changed.

Methodology

  1. Establish a baseline. Run a fixed set of 15 prompts and record mentions, citations, competitors, and answer patterns.
  2. Analyze competing answers. Study the products and sources surfaced for sewing planning, fabric stash, and pattern-matching queries.
  3. Make targeted changes. Improve first-party content and technical signals based on gaps revealed by the baseline.
  4. Rerun and document. Repeat the frozen prompt set, compare results, and publish what changed.

Baseline

The first frozen test establishes the starting point. These numbers do not claim improvement; they give future experiment rounds something consistent to compare against.

Baseline scorecard
ChatGPTClaudeGoogle AI ModeGemini
Prompt set
15 prompts per platform
Total responses
60
Brand mentions
0/60
Citation rate
0%
Phase 1

Build the AEO Foundation

After establishing a 0% AI visibility baseline, I focused on making Sew What’s Next easier for search and answer engines to discover, understand, and connect with its core use case.

What I implemented after the baseline

  1. Launched dedicated domain sew-whats-next.com.

  2. Created homepage that clearly described Sew What’s Next as a sewing planning app that matches fabric you already own with sewing patterns.

  3. Published a supporting page targeting the prompts being tracked.

  4. Added real product snapshots to demonstrate a functioning app.

  5. Added technical schema and fixed the sitemap to match the new domain.

  6. Verified crawlability, submitted the page and sitemap through Google Search Console, and began monitoring indexing.

Measurement 1

ChatGPT visibility reached 20%

After Phase 1, Sew What’s Next appeared in 3 of 15 ChatGPT responses. All three recommendations cited sew-whats-next.com.

ChatGPT visibilityUp 20 percentage points
ChatGPT20%3 of 15
Claude0%0 of 15
Google AI Mode0%0 of 15
Gemini0%0 of 15

ChatGPT visibility increased from 0 percent to 20 percent. Claude, Google AI Mode, and Gemini remained at 0 percent.

A note about the result: The increase happened after Phase 1, but this test cannot prove that those changes caused it. AI answers can change from one test to the next.

What I learned from Round 1

  1. Clear positioning helped ChatGPT find and cite the product. After I launched the website and added copy focused on what I wanted the app to be known for, ChatGPT surfaced it for specific prompts in a niche without a clear category leader.

  2. Publishing a website does not make it a trusted source. This matches my SEO experience: “If you build it, they will come” is not true. Google and Gemini did not cite the site simply because it was online. A new site still needs trust signals such as backlinks, external references, and domain authority.

  3. Other answer engines still favored established sources. Claude, Gemini, and Google AI Mode surfaced better-known products that have been around longer and appear in “best sewing app” articles. This matches AEO research I’ve read suggesting that Claude gives significant weight to third-party mentions from established media outlets.

Phase 2

Test an off-site signal

I contributed to an existing Reddit discussion about sewing tools. I kept the comment useful and non-promotional, and I did not link to Sew What’s Next.

Measurement 2

ChatGPT visibility reached 33%

Sew What’s Next appeared in 5 of 15 ChatGPT responses, up from 3 in the previous round. The recommendations stayed concentrated around the product’s core use case: deciding what to sew and matching fabric with patterns.

ChatGPT visibilityUp 13 points since Round 1
ChatGPT33%5 of 15
Claude0%0 of 15
Google AI Mode0%0 of 15
Gemini0%0 of 15

ChatGPT visibility increased from 20 percent to 33 percent. Claude, Google AI Mode, and Gemini remained at 0 percent.

A note about the result: I could not identify any visibility change caused by the Reddit comment. The overall ChatGPT increase may reflect stronger product association, normal answer variation, or other factors this test cannot isolate.

What I learned from Round 2

  1. A buried Reddit comment was not enough. My first test produced no observable change. The comment may have been too deep in the discussion, and without a link, the connection to Sew What’s Next may have been too weak.

  2. The frozen prompts help the experiment, but they are not the best long-term tracking set. Some prompts lead to general sewing advice instead of product recommendations. In a real AEO program, I would refine the set using actual search behavior, visibility opportunities, and business value. Here, I am keeping the prompts frozen so each round stays comparable.

  3. A competitor appears to be adopting similar AEO strategies. It is moving beyond inventory tracking toward predictive, AI-assisted project planning and publishing targeted content around that position. It now appears alongside Sew What’s Next for prompts where my app had been a leading answer, making it a priority competitor to monitor.

  4. AI visibility is still platform-specific. ChatGPT cited Sew What’s Next in 5 of 15 responses, while Claude, Google AI Mode, and Gemini remained at 0. This aligns with what I’ve observed throughout the experiment: those platforms appear to rely more heavily on third-party mentions and established credibility than on first-party website content alone.

Results over time

  1. Round 0 · Baseline

    ChatGPT visibility began at 0 of 15 prompts, or 0%.

  2. Measurement 1 · After Phase 1

    ChatGPT visibility reached 3 of 15 prompts, or 20%. Claude, Google AI Mode, and Gemini remained at 0%.

  3. Measurement 2 · After the first off-site test

    ChatGPT visibility reached 5 of 15 prompts, or 33%. The other platforms remained at 0%, and the Reddit contribution produced no observable lift.

Next: Test a source already being retrieved

For the next round, I’ll contribute a useful, relevant mention to a Reddit discussion that an answer engine already cites for one tracked topic. I’ll include a link only if it helps answer the question and the community rules allow it.