NewAnswers MCP: one question, one checked answer for your agent.See how it works

Product

AI competitive analysis: what each public source tells you about a rival

Updated

AI competitive analysis is only as good as the record it reads. A model can summarise a rival in seconds, but whether the summary holds depends on which public sources it used and whether anyone checked the claims against them. This page goes source by source through what a competitor reveals about itself and what each source cannot tell you, then gives an outline you can copy and says when a one-off sourced analysis is the right purchase and when a monitoring platform is.

What each public source reveals, and where it stops

Annual and quarterly reports: segments, customers and the company’s own metrics

A listed competitor’s 10-K and 10-Q are the most reliable numbers you will get, because they are filed under securities law. The FASB’s ASU 2023-07 requires a public company to disclose, annually and quarterly, the significant segment expenses its chief operating decision maker reviews, to name that decision maker’s title, and to give the full segment disclosures even if it has a single segment. It applies to fiscal years beginning after December 15, 2023. Regulation S-K Item 101 also asks for any dependence on key products, services or customers, which is where a rival admits that one channel partner or one government buyer carries its revenue.

The limit: the company defines its own metrics. Toast’s 10-Q defines gross payment volume and annualised recurring run-rate in its own terms, and a rival’s similarly named metric may be counted differently. Private companies file none of this. The EDGAR full-text search covers electronic filings since 2001; the SEC EDGAR API guide shows how to pull them in code, and account research from a 10-K shows which sections to read first.

For one number from a filing, a single question is enough. We asked Answers, the per-question tool:

Question: How many restaurant locations did Toast report at the end of its most recent quarter, and how fast is that number growing year over year?

Answer: Toast reported approximately 189,000 locations as of September 30, 2026, up 21% compared to one year ago.

Source: Toast’s Form 10-Q for the quarter ended September 30, 2026, quoted and checked word for word: “As of September 30, 2026, Toast served approximately 189,000 Locations, up 21% compared to one year ago, and processed $225 billion in gross payment volume over the trailing 12 months.” Check result: supported, with the quote verified. 13 sources consulted.

The competition disclosure

Item 101 asks filers to describe “trends in market demand and competitive conditions”, and smaller reporting companies their “competitive position in the industry”. A 10-K therefore often names the rivals the company sees, a cheap way to find ones you missed. Read it as the company’s framing. Toast’s latest 10-Q, for example, says only that it operates “in a very competitive and rapidly changing environment”. Share figures rarely appear.

Pricing pages and their archived history

The current pricing page shows list prices and how plans are cut. The history shows strategy, such as the quarter a free tier vanished. The Wayback Machine’s availability API returns the snapshot closest to any date you ask for, so you can line up a rival’s page by quarter. The archive itself warns that not every date for every site is complete, and that pages blocked by robots.txt, behind a password or excluded at the owner’s request are missing. Neither source shows negotiated discounts. Many enterprise vendors publish no price at all: Crayon’s pricing page offers a “Request an Estimate” form and Klue’s leads to a demo.

Patent filings and the 18-month lag

US applications are published 18 months after the earliest filing date, so a patent search shows what a rival was working on about a year and a half ago. The same statute lets an applicant who will not file abroad request nonpublication, and provisional and design applications are not published at all. The USPTO’s Patent Public Search covers granted patents and published applications. Search under every subsidiary the rival lists in Exhibit 21 of its 10-K, which Item 601(b)(21) requires, though subsidiaries that are not significant may be left off the list.

Job postings

Hiring shows where a rival is building. A study in Contemporary Accounting Research (Cao, Cheng, Tucker and Wan, 2023) found that firms under intense technological competition write more specific skill requirements into tech job ads, about 27 more specific skills on average, with a weaker effect at firms that rely on trade secrets. Pay is visible in some places: since November 1, 2022, employers with four or more employees advertising a New York City job must post a good faith salary range. A posting does not prove a hire, and a count of open roles is not headcount.

Court and enforcement records

Lawsuits surface disputed contracts and patent fights. PACER charges $0.10 a page, capped at $3 a document, and waives the bill if you spend $30 or less in a quarter. The FTC’s cases and proceedings library is searchable and filterable by industry, competition or consumer-protection topic and case status. A complaint is one side’s allegations until the docket shows an outcome.

Press releases and 8-Ks

Press releases are the company’s chosen story. The 8-K is the legal backstop: Form 8-K must generally be filed within four business days of a listed event, including material agreements, officer departures and material cybersecurity incidents. When the two differ, quote the filing. Our count of companies naming AI coding tools in SEC filings is an example of reading filings for a signal no press release would give.

A competitor analysis outline you can copy

  1. Question and scope. The decision the analysis serves, the rivals, the period, and what you will not cover.
  2. Company snapshot. Legal entity, subsidiaries, listed or private, last reported revenue by segment, with the filing cited.
  3. Customers and concentration. Disclosed customer dependence, segments served, the company’s own metric definitions.
  4. Product and pricing. Current list prices, archived prices by quarter, contract and cancellation terms.
  5. Direction. Patent filings (remember the lag), hiring by function and location, 8-K events of the last 12 months.
  6. Risks and disputes. Litigation, enforcement actions, recalls, disclosed risk factors that name you or your category.
  7. What the record does not show. Each gap stated plainly, with the estimate you chose not to make.
  8. Implications. Where you win, where you lose, what to watch next quarter.

Section 7 is the one most analyses skip. In the sample report Toast vs Square vs Clover, an unedited Deep run of about 7,500 words and 241 sources, Toast is the only one of the three that discloses restaurant location counts, so the report says a three-way location share cannot be built from filings instead of guessing one. For sizing the market the rivals share, see AI market research.

A one-off sourced analysis or an always-on platform

Crayon says it “monitors your competitors and alerts you to relevant intel” and delivers battlecards into Salesforce and Slack. Klue describes a Compete Agent that gathers competitive intel and gets it to sellers. Both are sold by quote. Buy one of these when a sales team meets the same rivals in deals every week and needs battlecards kept current. To learn only when a rival’s page changes, a page monitor is cheaper: Visualping has a free plan for five pages checked hourly and paid plans from $14 a month.

Buy a one-off analysis when the output is a document someone will act on, such as a board memo or a vendor choice, and the value is the citation behind each claim. Hyperresearch does this job. You name the rivals and what to compare; a run reads SEC EDGAR filings and XBRL financials, USPTO records, Wayback Machine snapshots, agency enforcement data such as EPA ECHO and Department of Labor records, and public pricing and job pages fetched logged out (the full list is on /sources). It does not pull PACER dockets, and it does not read Crunchbase, PitchBook or any page behind a login. Both tiers check cited claims against their sources and correct what does not hold; a Deep report adds a verification receipt with a verdict for each citation, and every source stays in your vault for next quarter’s refresh. The stages are on how it works. Hyperresearch does not monitor or send alerts. On the DRACO research benchmark, in our own grading, it scores 77.7.

Consultants doing this for clients can read AI for consultants; buyers assessing one target should read AI due diligence.

Starting a competitor analysis

A Light run keeps about 20 to 50 sources in about 30 to 40 minutes and suits one rival against you; a Deep run targets 150 to 200 sources over 3 to 5 hours and suits three to five rivals. Neither figure is a guarantee. Your first Light report from Hyperresearch is free when you sign up, with no card, at the console. Run sizes are on pricing. The analysis gathers and checks sources; it does not give investment advice.