We asked four AI assistants for a competitor monitoring tool 48 times. Here is who they named.

If you sell competitor monitoring software in 2026, your newest competitor is not another tool. It is the answer an assistant gives when a buyer asks "what should I use to keep an eye on my competitors". We wanted to know what that answer is, where it comes from, and how a product gets into it. So we counted.

Method

• When: 7 September 2026.
• Who: ChatGPT, Claude, Gemini and Perplexity, each with web search on, each asked in a United States context.
• What: twelve neutral buyer questions, no brand names in the prompt. Examples of the type: which tool tells me when a competitor launches new ads; how do I track a competitor's website changes every week; what should a small agency use to report competitor activity to clients. Four of the twelve are held out of all our own content planning.
• How counted: 48 answers. A language model separated a recommendation from a passing mention and listed the other vendors each answer put forward. A human read the classifications. Failed runs (none this time) would be excluded, not counted as zeros.
• Cost: $1.86 in API calls. The full panel with branded accuracy questions runs about $8.

The shortlist the assistants give

Vendors named as a recommendation, counted once per answer.

Three things stand out.

The assistants answer a category question with three categories. Ask for "competitor monitoring" and you get an SEO suite for the rankings half, a change monitor for the "tell me when something changes" half, and an enterprise CI platform if the question smelled like a team. Nobody in the top ten does the whole job for one business at a small-business price. That is a gap, and it is where we chose to build.

The names we expected barely appear. When we planned this, we assumed Competitors App, RivalHound and Competitaurus were the comparison set. Competitors App was named four times. The other two, never. The assistants do not know the small tools exist, because the pages that would tell them do not exist either.

Figo scored zero, and that is the useful number. On the day of the run the site had no sitemap, no structured data and no About page in any index. Every later run is measured against the same 48 answers, so the denominator is fixed.

The sources behind the answers

Cited domains across the 48 answers, most cited first. This is the list that matters if you want to change the answers.

Two thirds of the sources are competitor-owned blogs. The assistants are largely reading vendors' own "best competitor monitoring tools" articles and repeating them. Those pages will never list a rival, so they are not an outreach list. The practical list is short: HubSpot, Zapier, Buffer, Metricool, The CMO, RankMasters, AgencyAnalytics, and the two self-serve directories.

It also explains why an owned "best tools" article is worth writing even for a new product. It is the format the assistants demonstrably read for this category. Ours is here, with the other vendors treated fairly and the prices checked.

How often the assistants were wrong

The branded half of the panel asked sixteen questions about Figo by name. Those answers contained 39 statements about "Figo" that were false. Most described a pet insurance company with the same name: its Trustpilot reviews, its membership benefits, its claim disputes. A few described an unrelated recruiting product. Not one assistant described a competitor monitoring tool, because there was nothing to read.

The lesson is not specific to us. If your brand name is shared with a bigger company, the assistants will describe the bigger company, and your buyers will hear about someone else's claim disputes. The fix is entity work: a clear one-sentence description repeated everywhere, Organization markup with links to real profiles, an About page, and consistent naming. Brand search alone will be contested for a long time; "Figo competitor monitoring" is ours to win.

What a new product has to do

From this run, in order.

1. Get the pages crawled. Sitemap submitted, robots open to AI crawlers, structured data on every page.
2. Create the self-serve profiles the assistants cite: G2, Capterra, LinkedIn, Crunchbase, and link them from the Organization markup.
3. Write comparison pages against the tools buyers are actually being told to consider: Semrush, Visualping and Crayon, before anything else.
4. Pitch the eight independent sources with one honest paragraph each.
5. Re-run the same twelve questions in 30 days. Add other countries once the US number moves off zero.

Limits

Forty-eight answers is a starting line, not a market study. The counts are shares of this fixed panel, not market share, not search volume, not evidence of demand. One run per question means differences of a few answers between tools are not meaningful; only the zero result and the top three are unambiguous. The raw answers and classifications are kept so the next run can be compared line by line.

This is the same measurement Figo runs for its customers every Monday: ten buyer questions in their niche, four assistants, recommended and cited counted separately, sources listed. How to run it for your own business.