Pineprompt

Track what ChatGPT, Gemini, Perplexity and Google AI say about your brand. Building in public: real numbers, AI search data, what we learn.

[Pineprompt @pineprompt]

Google AI Mode splits one question into several searches before writing its answer. You do not compete for one query, you compete across the cluster. Map the sub-questions and make sure you have a page that answers each.

[Pineprompt @pineprompt]

Category matters for volatility. In our 14-day benchmark, CRM prompts changed answers day to day 48.9% of the time. If you sell in a crowded category, weekly checks are not enough.

[Pineprompt @pineprompt]

Pineprompt is launching on Fazier today. It tracks how ChatGPT, Gemini, Claude, Google AI Overviews, AI Mode and Copilot mention your brand, every day, with the full answers and cited sources.

https://t.co/7y2ASeU7IX

[Pineprompt @pineprompt]

8,269 distinct brands appeared across 18,792 AI answers in our benchmark, 6.2 per answer on average.

The list AI recommends from is long, and it shifts by engine and by day.

[Pineprompt @pineprompt]

One AI answer is a sample of one. In our Sep 30 to Oct 13 benchmark, Perplexity's answer to the same prompt changed on 68.8% of days. Run every prompt several times before drawing a conclusion.

[Pineprompt @pineprompt]

Being mentioned and being cited are different results. A brand can be named with no link to its site, or linked without being recommended. Track both as separate columns, or you will misread what changed.

[Pineprompt @pineprompt]

Tip for tracking Gemini: log whether each answer was grounded with sources or answered from memory. If mentions drop only in ungrounded answers, it is training data. If they drop in grounded ones, look at which pages got cited instead.

[Pineprompt @pineprompt]

Copilot named 7.6 brands per answer on average. Perplexity named 3.9.

Some engines leave much more room to be mentioned than others, which changes how hard each one is to win.

[Pineprompt @pineprompt]

We republished our 7 engine tracking guides (ChatGPT, Claude, Gemini, Perplexity, Copilot, AI Overviews, AI Mode) on and Hashnode, each with a canonical link back to the original. Same content, more places to be found, no duplicate content risk.

[Pineprompt @pineprompt]

Most "AI SEO" advice is regular SEO with a new name. What's actually different: AI engines lean on lists, reviews and forums, not just your own site.

[Pineprompt @pineprompt]

Build in public, week update: 546 pageviews in the last 7 days, 0 paying customers so far. Biggest lesson: cold email to agencies has not converted. This week we are moving effort to answering real buyer questions where they are asked.

[Pineprompt @pineprompt]

In our Sep 30 to Oct 13 benchmark of 200 prompts, Perplexity cited 349 distinct domains. ChatGPT cited 190.

Perplexity spreads its trust wider, which gives smaller sites more ways in.

[Pineprompt @pineprompt]

We tracked 200 buyer prompts across 7 AI engines for 14 days. 8,269 brands came up.

46.2% of them were named by only one engine. Just 12.9% were named by all seven.

Checking one engine tells you very little about your AI visibility.

[Pineprompt @pineprompt]

Google AI Overviews and ChatGPT agreed on only 45.6% of the brands they named for the same prompts.

Ranking in one AI surface does not carry over to another.

[Pineprompt @pineprompt]

Building in public: our Google search impressions went from ~400/day to ~120/day after we retitled one page. The old title ranked top 10 for a query we cared about.

Lesson: don't touch pages that are working. Watching it a few more days before reverting.

[Pineprompt @pineprompt]

AI answers move fast. On Perplexity, 68.8% of the brands named for a prompt changed from one day to the next in our 14-day benchmark.

A single screenshot of an AI answer is not a measurement. Track the same prompts daily.

[Pineprompt @pineprompt]

If an AI answer cites a "best X tools" list and you're not on it, that list is your outreach target. Not the AI.

[Pineprompt @pineprompt]

Dataset and CSV: https://t.co/ZRLHhgmPC1

[Pineprompt @pineprompt]

We published our AI search benchmark under CC BY 4.0, with the CSV. 18,792 answers, 200 prompts, 10 categories.

Use it, chart it, argue with it:

[Pineprompt @pineprompt]

What moves AI answers, from what we see in cited sources: comparison listicles, review sites, Reddit threads, and your own pages with clear answers.

Not keyword density.

[Pineprompt @pineprompt]

@faraujoj Gracias por compartirlo, Fabio. El dataset completo (CSV, CC BY 4.0) está aquí por si alguien quiere revisar su categoría: https://t.co/ZRLHhgmPC1