How to Benchmark Your AI Readiness Against Competitors

Your AI readiness score on its own tells you almost nothing. It tells you where you are. It doesn't tell you whether that's good. If your whole category is asleep at the wheel, a mediocre score still puts you ahead. But if your top three competitors have already published llms.txt files, structured data, and open API endpoints, then nailing one check doesn't close the gap. It barely dents it.

Benchmarking against competitors is fast becoming standard practice for growth teams and agencies juggling multiple brands. Here's how to actually do it, what to measure, and what to do once the numbers are in front of you.

Why Benchmarking Matters More Than Absolute Scores

AI models don't recommend businesses in a vacuum. They recommend you relative to everyone else in your category. When someone asks “what are the best tools for project management?”, the model builds a list from what it knows about the whole field. The names that surface are the ones with the strongest mix of training-data presence, structured signals, and entity verification, measured against every other business competing for the same slot.

And in the short to medium term, this is a zero-sum game. If a model returns five recommendations, every new entrant that earns a spot knocks an existing one out. So your goal isn't just a better absolute score. It's to be more legible to AI than the businesses sitting next to you on that list.

What to Benchmark

A complete competitive AI readiness benchmark covers four dimensions:

1. Technical signals

These are measurable by scanning each competitor's site. The key signals to check:

• AI crawler permissions (robots.txt): are GPTBot, ClaudeBot, PerplexityBot allowed or blocked?
• llms.txt: does it exist, and how detailed is it? (The spec lives at llmstxt.org if you want to see what a good one looks like.)
• JSON-LD schema: is there Organization or Product schema? What fields are populated?
• sameAs links: how many external entity profiles are linked?
• agents.json: does it exist?
• Sitemap: is it published and referenced from robots.txt?
• Content-type headers: are they correctly configured?

These signals are publicly visible and can be scanned for any domain using Faro's AI Readiness Scan. Run your competitors through the same scan you run on yourself and compare scores directly.

2. AI mention rate

The mention rate is the percentage of AI model queries in your category where a given business appears. It's the most direct measure of AI visibility and the one that most closely maps to business impact.

To benchmark mention rates, you need to run the same set of category queries across AI models for both your business and your competitors. A standard benchmark would include queries like:

• “What are the best tools for [category]?”
• “Recommend a [category] solution for [persona].”
• “Compare the leading [category] platforms.”
• “Which companies offer [specific feature or capability]?”

Run each query across ChatGPT, Claude, Perplexity, and Gemini. Record whether each competitor appears in each response, and what position they appear in. This gives you a cross-model mention rate and a position distribution for each competitor.

Faro's Competitor Intelligence tooling does this automatically. You set your business and up to two competitors, and it runs the queries across all four models and hands back a share-of-voice comparison. No spreadsheet wrangling required.

3. Sentiment quality

Not all mentions are equal. A competitor that appears in AI responses but is described as “expensive and complicated” is effectively getting negative advertising. Benchmark not just whether competitors are mentioned, but how they're described.

When reviewing AI responses, look for:

• How is each competitor positioned (leader, alternative, budget option, enterprise choice)?
• What features or strengths does the AI attribute to each?
• Are there consistent negative qualifications (e.g., “though it can be pricey”, “steep learning curve”)?

Sentiment benchmarking reveals opportunities: if a competitor is frequently mentioned but with consistent negative qualifiers, and you don't have those qualifiers, you can emphasize the contrast in your own content and schema.

4. Entity strength

Entity strength measures how well-established a business is as a verified entity across the web. Indicators include:

• Wikidata entry existence and completeness
• LinkedIn company page with accurate data
• Crunchbase profile
• Number of sameAs links in schema
• Mentions in industry publications with links
• Google Knowledge Panel presence

This one's harder to put a number on. But you can do a quick qualitative audit: search each competitor's brand name in Google, then look at the Knowledge Panel, the Wikidata entry, and the general quality of third-party coverage. Take a B2B HR software company sizing itself up against Rippling and Gusto. Both of those incumbents have rich Knowledge Panels, complete Wikidata entries, and years of press. The smaller player almost certainly doesn't, and that gap is exactly the kind of thing a benchmark surfaces in black and white.

Building a Competitive Benchmark Report

A practical competitive benchmark for AI readiness has four components: a technical signal matrix, a mention rate comparison, a sentiment summary, and a priority action list.

Technical signal matrix

Create a table with competitors as columns and the key technical signals as rows. Mark each as Pass, Partial, or Fail. This gives an immediate visual read of where each competitor stands and where you have gaps or advantages.

Example columns: Your Brand, Competitor A, Competitor B, Competitor C. Example rows: AI crawlers allowed, llms.txt published, Organization schema, Product schema, sameAs links, agents.json, sitemap. Each cell is a simple status.

Mention rate comparison

Run five to ten representative category queries across four AI models. For each query, record which businesses appear. Calculate a mention rate for each business as mentions divided by total possible mentions (queries times models). Express as a percentage.

A mention rate of 40% means a business appeared in 40% of all query/model combinations tested. If your rate is 25% and your top competitor is at 60%, that's a 35-point gap to close.

Sentiment summary

For each competitor that appears in AI responses, note the most common descriptors and qualifiers used. Look for patterns across models. If three of four models describe a competitor as “the enterprise option”, that's their de facto AI positioning, regardless of how they market themselves. If they're consistently described as “complex to set up,” that's a vulnerability you can address in your own positioning.

Priority action list

The output of a competitive benchmark isn't a score. It's a prioritized list of actions. Good actions are those that (a) close a gap where a competitor has an advantage you lack, or (b) extend a lead you already have. The signal matrix and mention rate comparison together identify both.

How Often to Benchmark

For most businesses, a quarterly competitive benchmark is appropriate. AI model knowledge updates on roughly 3 to 6 month cycles for foundational training data, and market conditions in most categories don't shift faster than that.

Exception: if your category is rapidly growing or you have a major competitor that is actively investing in AI readiness, monthly benchmarking is warranted. The first mover advantage in AI recommendation visibility is real, and the gap between early movers and laggards compounds quickly.

For agencies managing multiple clients in competitive categories, Faro's continuous tracking watches mention rates and alerts you when something shifts, which replaces the manual quarterly check entirely. You can see which tier covers competitor tracking on the pricing page.

What to Do with the Results

If you're behind on technical signals

Technical gaps are the fastest to close. Unblocking crawlers, publishing llms.txt, and adding JSON-LD schema can be done in a day or two, and they land immediately on models with live retrieval. Do these before anything else. Honestly, the fact that they're this cheap to fix (which, in most categories, is still pretty rare) is why a laggard can leapfrog a slow incumbent in a single afternoon.

If you're behind on mention rate despite good technical signals

This suggests a training data gap. Your competitors have more historical presence in the data these models were trained on. The fixes here are longer-term: consistent publishing, earning press coverage and citations, appearing in industry roundups and comparisons, and building third-party content that references you alongside competitors in your category.

If your sentiment is weaker than competitors

Sentiment in AI responses correlates with how your business is described in the sources those models were trained on. If you have consistent negative qualifiers, look at where those qualifiers come from: review sites, comparison posts, forum discussions. Addressing the underlying issue (pricing, complexity, support) is the root fix. In the shorter term, creating authoritative content that provides balanced, accurate descriptions of your product can shift the signal over time.

The Benchmark as a Business Case

For agencies, this is one of the easiest retainers to sell. The format is already familiar, since it mirrors the competitive SEO audits agencies have presented for years. The data is objective. And the output is something a client can act on. Show someone that their competitor sits at a 45% mention rate while they're stuck at 15%, and they get it instantly, long before anyone has to explain what AEO stands for.

In Short

An AI readiness score only tells you half the story. The competitive benchmark tells you the other half: whether your score is good enough to beat the businesses you compete with for AI model recommendations. Run technical signal scans on your top competitors, measure mention rates across AI models, and track sentiment quality. The output is a concrete set of actions prioritized by competitive impact, not just absolute score improvement.

Start by scanning your competitors with Faro's AI Readiness Scan. Then compare your technical signals side by side and identify your highest-leverage gaps.

Frequently Asked Questions

Is it ethical to scan competitors' websites?

Scanning publicly accessible signals on competitors' websites (robots.txt, structured data, publicly declared files like llms.txt) is standard competitive intelligence, no different from manually reviewing a competitor's website. The signals we're examining are designed to be publicly accessible. Faro's scan does not access protected or non-public content.

How accurate are AI mention rate benchmarks?

AI model responses have variance: the same query can produce different results on different runs, especially on models with live retrieval. A reliable benchmark requires running each query multiple times and averaging results. Faro's AEO Citation Monitor runs each query three times per model to account for variance and reports a stabilized mention rate.

Can a small business compete in AI recommendation visibility against large incumbents?

Yes, especially in niche categories. AI models tend to surface the most authoritative sources for a specific query, not just the largest brands overall. A small specialist in a well-defined niche with excellent structured data and a strong llms.txt can outperform a large generalist on queries where the niche is the focus. The niche specificity is the competitive advantage.