TL;DR
When a user asks an AI “which is better, you or your competitor,” the AI wants content where both sides appear in a single source — and most brands never write that page, because they’re afraid to name competitors. The result: comparison answers are decided almost entirely by third-party reviews and forum threads. This article covers how to map the four shapes of competitor queries, the content specs for three page types (comparison page / alternatives page / use-case selection page), and the measurement loop after launch.
The Question You’re Absent From
Open ChatGPT and ask “which is better, [your brand] or [your biggest competitor].” You’ll usually get one of two outcomes: you’re not mentioned at all, or — worse — the AI puts you in a comparison table, but every spec, price, and pro/con was pulled from a forum thread or a third-party review, and you had no say in any of it.
This isn’t an isolated glitch. McKinsey’s 2025 report on AI search (New front door to the internet) contains one key figure: brand-owned pages account for only 5–10% of AI search citations. More than nine out of ten citation slots go to third-party content — and on comparison queries specifically, brands are absent even more thoroughly than average, for plain reasons:
- When an AI builds a comparison table, it prefers a single source that already covers both sides. Your website only talks about you, which makes it structurally unfit for answering “A vs B.”
- Brands are afraid to name competitors. Legal plays it safe, sales hates giving rivals airtime, and the whole query surface gets ceded to review sites and forums.
- Your competitors are usually just as afraid. That makes comparison content one of the few asymmetric, first-mover opportunities left — the official version you publish is often the only first-party source for this entire question.
Third-party versions won’t disappear, and they shouldn’t. Your job is to make sure that when an AI generates a comparison answer, there’s one version on the table that you actually fact-checked.
Map the Query Surface First: Four Shapes of Competitor Keywords
Before writing any page, map how users actually phrase the question. Competitor-related AI queries fall into roughly four shapes:
| Shape | Example | Where the user is stuck | Page type |
|---|---|---|---|
| Direct comparison | “A vs B — which is better?” | Shortlisted two or three options, wants a verdict | Comparison page |
| Alternatives | “alternatives to A” “I don’t want A, what else is there” | Unhappy with the status quo, wants a list | Alternatives page |
| Use-case selection | “which X suits a small business” “what should I pick on a $1,000 budget” | No brand shortlist yet, searching by conditions | Use-case selection page |
| Switching / migration | “what to watch out for when moving from A to B” | Already decided to switch, afraid of migration pain | Comparison page’s FAQ section |
To build the list, reuse the 10-minute self-test from the earlier article: combine your category terms, brand name, and main competitor names into the four sentence shapes above, run them through ChatGPT, Gemini, Perplexity and other mainstream AI engines, and record three things — whether you’re mentioned, who gets cited, and whether the description is accurate. This list drives your layout, and doubles as the baseline for post-launch measurement.
Four query shapes don’t require four kinds of pages. Three page types cover them:
Page Type 1: The Comparison Page — One Rival Pair Per Page, Verdict First
The comparison page answers “A vs B.” It has the strictest structural requirements of the three, because the AI will dismantle your comparison table and rebuild it inside its own answer. The spec:
| Section | Purpose | Spec |
|---|---|---|
| Opening verdict | Give the AI a conclusion it can grab | First paragraph states “pick us when…, pick them when…” in under 40 words |
| Spec comparison table | For row-by-row reuse by the AI | Fill each cell with a short phrase, not a checkmark; write prices, limits, support scope as concrete values |
| Who each one suits | Catches use-case follow-ups | 2–3 scenarios per side, written as conditional sentences |
| Where the rival wins | Trust signal | At least one specific, concrete concession — no polite filler |
| Migration FAQ | Catches switching-intent queries | One question each on cost, timeline, and data migration when switching from the rival |
| Verification date | Credibility and maintenance anchor | State “competitor information verified as of YYYY-MM,” update on every rival release |
A few common mistakes.
Every table cell must stand alone as a sentence. AIs extract tables cell by cell. A “✓” or “✗” carries no information once extracted; “API export supported (JSON/CSV)” or “paid plan only” travels intact. This is the same principle as content citability, applied to tables.
The row where you concede to your rival is the most valuable row on the page. The reason is purely practical: the AI’s goal when answering a comparison question is to help the user decide. Content that lists only one side’s strengths gets classified as advertising and skipped; content that concedes “they’re genuinely the better fit for scenario X” earns the standing of a neutral source. You give up one row and buy citation eligibility for the whole page.
One rival pair per page — don’t build a “us vs. everyone” mega-page. Queries arrive in pairs (“A vs B”), so pages should answer in pairs. Three competitors means three pages.
Page Type 2: The Alternatives Page — Listing Only Yourself Equals Not Writing It
The alternatives page answers “what are the alternatives to X.” The X splits into two situations with completely different strategies:
X is the market leader and you’re the challenger — this is an offensive page. Write “N alternatives to X,” list 4–6 options (including yourself), and give each an honest one-paragraph positioning: who it suits, where it’s strong, what its obvious gap is. Write your own entry in exactly the same format as everyone else’s, with no adjective barrage — visual equality on the page is what buys credibility. An “alternatives page” that lists only you is an ad, and an AI won’t use an ad to answer a list-type question.
X is you — this is a defensive page, and it matters more than most teams assume. Someone asking “alternatives to [your brand]” has churn intent already. If you don’t answer this question, the answer gets assembled entirely from competitors’ offensive pages and forum complaint threads. Publish your own “[your brand]: alternatives and when each fits,” honestly listing rivals — while spelling out switching costs, data-export limits, and the gaps you’ve recently closed — so that someone considering leaving at least reads one version you participated in.
Both situations share one spec: every claim about every vendor must be verifiable. Only use competitor prices and feature limits that public sources confirm, and state the verification date. Getting a rival’s spec wrong costs the credibility of the entire page.
Page Type 3: The Use-Case Selection Page — Catch the People With No Shortlist Yet
The use-case selection page answers queries like “which X suits a small business” or “what should I pick on a $1,000 budget” — questions that name no brand at all. It carries the highest query volume of the three page types and is the least sensitive to write: you’re not dueling anyone, you’re shaping selection logic into a form an AI can carry.
The core format is the conditional sentence: “If you are [scenario], choose [solution type], because [reason].” The whole page is a set of conditionals plus one decision table:
If your team has no engineers → choose a fully managed service; the maintenance cost of self-built tooling eats whatever it saves
If you publish 20+ pieces of content a month → choose a solution with batch analysis; manual page-by-page checks won't survive month two
If you just want to verify there's an effect → run a free scan first, then decide whether to commit budget
When an AI cites this kind of page, it frequently lifts entire conditional sentences into its answer — reasons included. Two spec points: the conditions must be mutually exclusive and cover the main scenarios (every scenario you skip is someone else’s citation opportunity), and honestly state when someone should not pick you. Same logic as the comparison page: you give up one cell to buy neutrality for the whole page.
Writing Specs: Extractable for the AI, Defensible for Legal
The underlying specs shared by all three page types, each with a concrete reason:
Conclusion first in every paragraph. The first sentence of each paragraph is its verdict; evidence follows. AIs weight lead sentences heavily when summarizing — a conclusion buried in sentence three may as well not exist.
Numbers must be traceable. Princeton’s GEO research (Aggarwal et al., KDD 2024) measured this: adding verifiable statistics to content lifts visibility in generative engine citations by roughly 40%, and adding sourced quotations by about 28%. The inverse also holds — an unsourced number is a liability. “Industry consensus says” loses to “per the 2026 X report.”
Cross-page consistency. The prices, plan names, and feature scopes on your comparison page must match your pricing page, FAQ, and homepage exactly. AIs cross-read multiple pages of yours, and conflicting versions of brand facts sink the credibility of both pages at once.
The legal red line on competitor information. In Taiwan, comparative marketing that references competitors falls under the Fair Trade Act, and false or misleading comparisons can draw penalties. Operationally, hold three rules: use only the rival’s public information, keep a verification record for every spec (screenshot or archived link), and print the verification date on the page. That date doubles as a freshness signal for GEO.
Don’t stack comparative adjectives. “Faster, more stable, better value” is zero information to an AI and pure risk to a regulator. Rewrite every “more” as a verifiable statement.
After Launch: The Measurement Loop
Comparison content is alive — a rival ships a redesign, changes pricing, or renames a plan, and your page is stale. Finishing the layout is the start, so run a fixed loop:
- Baseline: before writing, run your query list through each platform once and keep the records.
- Re-test at two weeks and one month after launch: same queries, same record format — mentioned or not, who got cited, whether the description is accurate. A single run proves nothing; AI answers exhibit sampling drift, so sample the same query across multiple rounds and read the trend.
- Rival releases trigger updates: when a competitor changes pricing or features, your comparison page updates within the month, verification date included. Stale comparison content is worse than none — it makes the AI cite a wrong version of you.
The manual procedure is fully documented in the manual backtesting guide; once the query list grows and the platform count multiplies, manual backtesting hours spiral — which is the main reason most teams’ layouts don’t survive month two.
Caveats
Three things, stated once. The size of the competitor query surface varies enormously by industry — a B2B SaaS and a local service business have completely different shape distributions, so test before you lay out pages, and don’t copy someone else’s page-type ratio. The specs for the three page types are a skeleton; fitting them into your CMS involves templates, schema, and cross-page architecture that this article deliberately leaves out. And comparison content in YMYL industries such as finance and healthcare carries extra compliance requirements — clear legal review before writing.
Do two things first: this week, use the 10-minute self-test to map your competitor query list; then pick the rival pair where you’re bleeding worst and write the first version of a comparison page to the Type 1 spec.
Everything after the layout — multi-platform mention tracking, update cadence on rival releases, cross-page consistency maintenance — is the long-haul work geoweb.tw’s managed service exists for. 👉 Run a free GEO scan to see where you stand across 12 dimensions; if you want someone to carry the long-term upkeep, write to [email protected].
Part of the GEO technical deep-dive series. Further reading: Your Competitors Are Already Recommended by AI — Are You?, Content Citability, The Truth Behind AI Citation “Drift”