Find where competitors are preferred over your brand

This playbook helps you narrow a broad competitive gap to a specific customer need, AI system, or decision context. You will then decide whether the right response is correcting information, improving evidence, addressing a product limitation, or gathering more data.

Your outcome: one or more scoped competitive gaps, each with supporting evidence and a justified next action.

Choose the symptom you recognize

Start with one gap that matters to your business. You can finish this playbook with one well-supported task; a complete competitor audit is not required.

Before you begin

Use a relevant Context with observed results in Prompt Universe or an AI Visibility Benchmark. If the brand is absent and you have not yet established why, start with Find out why your brand is missing from AI answers. If your benchmark is ready but has not collected observations, start observing before interpreting performance. A ready setup is not a measured result.

Choose the competitor you actually need to compare with. A familiar market leader may be less relevant than the provider competing for the same audience and use case.

For this playbook, focus on discovery and recommendation situations. Questions that explicitly ask about a competitor belong to a different task: assessing whether your brand is introduced as an alternative.

1. Make sure the comparison answers the same question

Confirm the following before reading the gap:

For organic visibility, use questions that do not seed your brand. Keep controlled brand questions separate from discovery questions.

If comparable data is unavailable, investigate the missing scope or establish a consistent benchmark. Do not name a winner from mismatched evidence.

2. Start with the gap SpyderBot has identified

Open Decision Intelligence to review the contexts in which your brand wins or loses. If you are working from a benchmark, review Current Gap Breakdown by intent, AI system, or cluster.

Use the system's findings and evidence indicators to select a segment. Prefer a customer need you serve, a gap supported in the displayed scope, and a decision your team can act on. A large gap in an unrelated category is a weaker starting point than a smaller gap affecting an important buying requirement.

If all relevant findings have limited evidence, select one uncertainty to investigate. You do not need to inspect every question behind the overall result.

Look for patterns such as:

• A competitor is preferred for a particular requirement.
• Your brand appears during discovery but is rarely recommended during selection.
• A gap is concentrated in one AI system.
• A relevant cluster has too little evidence to support a decision.

Current Gap Breakdown describes the latest comparable observation. It does not, by itself, show that a gap is growing. Use comparable historical observations when the question is about change over time.

3. Read the answers that explain the selected gap

Start with the explanation SpyderBot provides for the selected gap, then open the supporting evidence needed to validate your decision. Examine responses that show the pattern and, where available, examples that do not follow it. Your goal is to test the explanation, not manually rebuild the analysis from every response.

For each relevant response, ask:

1. What did the customer need, including constraints?
2. Was each brand merely mentioned, included as an option, or actively recommended?
3. What reason did AI give for preferring an option?
4. What did it say about your brand's strengths or limitations?
5. Which sources were cited, if any?

A competitor appearing in a comparison table is not automatically the preferred choice. Likewise, your brand being mentioned does not mean AI recommends it for the customer's need.

For example, "A and B both offer invoicing; B is the better fit when automatic reconciliation is required" supports a preference for B under that requirement. It does not establish that B is preferred for every invoicing task. "A and B are options to consider" supports inclusion of both, without establishing a preference.

Use Explorer when you need a focused current investigation, such as testing an under-observed requirement or checking whether a pattern is specific to one system. Preserve the Context in the handoff. An Explorer investigation is a current snapshot; it does not create benchmark history.

4. Check the explanation against reality

The reasons in AI responses are evidence of what AI said. They are not automatically accurate descriptions of your product or proof of what caused the recommendation.

Check the relevant capability, pricing condition, availability, or limitation against an authoritative source for your own product. If competitor facts are material to the decision, verify those too.

Then classify the gap:

Do not assume publishing another page will solve every competitive gap. If the product does not satisfy the customer's requirement, more content cannot make that claim true.

When do you have enough to assign a task? You can state the affected customer need, point to relevant observed evidence, verify the fact or asset to change, and define an acceptance check. A single demonstrably outdated page can justify a factual correction. A claim that a competitor consistently wins requires broader comparable observations. Let the size of the decision determine how much validation you need, alongside the product's evidence indicators.

5. Choose a change you can justify

Turn the diagnosis into one concrete task. Specify the asset or product area, the evidence behind the task, and what the change should make clearer or better.

For example:

"Update the integration page to explain which plans support two-way sync, with setup requirements and a verified example. AI responses in the small-business integration cluster currently describe the integration as export-only."

This is more actionable than "Improve our AI visibility." It also gives the content owner a factual acceptance criterion.

Where citations are relevant, inspect them in Citations. Distinguish your own sources, competitor sources, and third-party sources. An observed citation tells you that a source appeared in the answer; it does not prove that changing that source will change the recommendation. If only partial source content is available, record that limit.

6. Define how you will check the result

First verify the change itself: the facts are correct, the page is accessible, and the intended information is present. Then review later comparable AI observations for the same customer need and metric.

Keep the comparator and scope consistent. If you change the question panel, market, or system selection, record the difference and avoid presenting it as a clean before-and-after result.

An improvement after a change is useful evidence of an outcome. It does not establish that the change alone caused the improvement. Other sources, competitors, and AI behavior can also change.

If the next observation still looks the same

Check the next eligible comparable observation in your monitoring program rather than assuming the next calendar day supplies a valid comparison. If the changed information is still portrayed incorrectly, verify the publication and inspect whether the observed answers rely on an older or different source. If portrayal is now accurate but the competitor is still preferred, revisit the stated selection criteria: the remaining gap may concern product fit or supporting evidence. If data is insufficient, retain the open question and specify which observation would resolve it.

Avoid repeating the same content edit merely because the first later answer has not changed. Record a review date appropriate to your monitoring schedule; no fixed waiting period guarantees that an AI system will update its answers.

Example: an apparent product gap becomes an information task

Illustrative example. A team sees a competitor preferred in questions about invoicing tools that connect to accounting software.

The team narrows the result to small-business integration questions in one AI system. Supporting answers describe its product as requiring manual exports. The product team confirms that two-way sync is available, but the public integration page still describes the older workflow.

The team records the following decision:

The team does not claim that it loses across all AI systems. It also does not promise that correcting the page will make it the top recommendation.

Record your competitive gap

You are done when the gap is specific enough to act on-or when you have identified exactly what evidence is missing. "No justified change yet" is a valid conclusion when the comparison is weak or the competitor is better suited to a need outside your target market.

Continue with: Fix issues affecting important pages, if the supporting information is present but a page-access or readability problem needs investigation.