SpyderBot Documentation
Improve how AI describes and positions your brand
AI may describe your brand positively while missing the advantage that matters to customers. It may also repeat an obsolete limitation, confuse your company with another entity, or place your product in the wrong category.
Your outcome: a specific portrayal issue, verified against reliable product information, with an appropriate correction or positioning task.
Start with the problem you recognize
| What you see | First question |
|---|---|
| "AI gets a fact wrong." | What exact claim is wrong, and what verifies the correct fact? |
| "AI misses our main advantage." | Is the advantage relevant to the question and supported publicly? |
| "AI describes us negatively." | Is the statement inaccurate, an opinion, or a real limitation? |
| "Different systems describe us differently." | Which scope contains the pattern that matters to the customer decision? |
Use AI Brand Portrayal, Brand Insights intelligence, a relevant Explorer investigation, or Brand Understanding observations in a benchmark. Start from the representation summary already available. You do not need all four sources.
1. Select one claim or attribute worth addressing
Look beyond the overall sentiment to category, positioning, strengths, limitations, and recurring claims. Choose a statement that affects a real customer decision: supported markets, integration capability, target audience, pricing conditions, or suitability for a use case.
Check which systems, questions, and observations support the pattern. A description in response to a question naming your brand is a controlled brand-understanding observation. Keep it distinct from how your brand is portrayed when AI introduces it in an unseeded category answer.
2. Verify what AI said and whether it is accurate
Open the supporting response. Read the exact wording and conditions. "May be less suitable for very large teams" is different from "does not support teams." Verify the underlying fact against current approved product information.
Classify the issue before choosing a response:
| Type | Appropriate response |
|---|---|
| Incorrect or obsolete fact | Establish the correct information and identify inconsistent public assets. |
| Accurate but unfavorable limitation | Explain the real trade-off or involve Product; do not publish an untrue correction. |
| Relevant advantage is missing | Check whether the question calls for it and whether clear public evidence exists. |
| Entity confusion | Check identity, names, descriptions, and structured information; investigate entity understanding where relevant. |
| Subjective judgment | Inspect the stated reasons and supporting evidence rather than declaring the opinion factually wrong. |
| Claim cannot be verified | Keep it unresolved and assign a fact-check. |
For one concrete factual error, a verified source may justify a correction. For a broad repositioning project, look for a supported pattern across the relevant observations before changing the message everywhere.
When the facts are correct but the positioning is incomplete
Choose the customer requirement that makes the missing advantage useful. Compare the supported positioning in that scope with the position your team wants to communicate. Identify the evidence a customer could use to evaluate the advantage: actual conditions, a demonstration, approved product details, or a relevant documented case.
For example, if AI accurately describes your footwear range but omits its fit options in questions about wide feet, check that the options are available in that market and clearly documented. A supported omission can justify improving the fit information. It does not justify claiming that every customer will find the shoes comfortable. Track the relevant attribute and use case, not only overall positive sentiment.
3. Find the information your team can change
Inspect the relevant owned pages and any cited sources. Distinguish the AI answer's portrayal from what an accessible source actually says. A citation beside a claim does not establish that the source contains or caused that claim.
If your product page and help center disagree, align them around a verified fact. If content is accurate but retrieval or entity interpretation appears problematic, use the relevant GEO Audit investigation. If a third-party article is outdated, prepare a correction request with the exact statement and authoritative supporting information.
When source content is unavailable or only partially accessible, keep that limit in the diagnosis. Do not attribute the whole portrayal problem to a page you have not inspected.
4. Write the smallest useful correction or positioning brief
Illustrative example. AI describes a footwear brand as "primarily formal shoes" in questions about comfortable commuting. The brand now offers a commuting range, but its category description and older retailer listings still emphasize formal wear.
The team first verifies the range and comfort claims. It chooses to update the owned category description and provide current product facts to the relevant retailer. It does not try to replace every mention of formal shoes: that remains a valid part of the brand.
Observed portrayal: primarily formal shoes in the selected commuting context.
Verified reality: a commuting range exists; use only approved product attributes and supporting evidence.
Task: clarify the range, intended use, and limitations on the owned category page; prepare factual information for the retailer.
Owner: brand content, with product approval.
Acceptance: current range and claims are represented accurately across the targeted assets.
Later check: review portrayal in comparable commuting questions; track recommendation separately.
5. Review the specific portrayal after implementation
Use later comparable observations to check the targeted claim or attribute. If it is now accurate but sentiment remains mixed, examine the remaining limitations rather than assuming the correction failed. If one system continues to repeat the old description, investigate that scope and its available source evidence.
Keep the outcome unresolved when observations are insufficient. You can improve public information and check observed answers; you cannot directly rewrite an AI system's knowledge through SpyderBot.
You are done when the portrayal issue is classified, the relevant facts are verified, and the next task has an owner and a targeted outcome check. Keeping an accurate limitation unchanged is a valid decision.
Next: use Turn an AI visibility opportunity into a content brief for an owned asset, or Choose which sources and channels to improve first for external information.