SpyderBot Documentation
Decide which AI questions to investigate and monitor first
SpyderBot has generated questions for your brand, but your team cannot investigate every question or monitor every possible customer need. Where should you start?
Use this playbook to turn the recommendations in Prompt Universe into a focused set of next steps. You will decide which questions need fresh evidence, which already support an improvement, and which are worth tracking over time.
Your outcome: a selected set of customer needs, a reason for including each one, and a next action: investigate, improve, monitor, or defer.
Start with the decision you need to make
| Your situation | Start here | Stop when |
|---|---|---|
| "I have recommendations but do not know where to begin." | Review Next Best Actions in step 2, then use the action table in step 5. | A relevant recommendation has a confirmed next action. |
| "The suggested questions do not reflect our customers." | Check the Context in step 1, then coverage in steps 3-4. | The missing customer need is represented without unnecessary duplicates. |
| "I need to track our competitive position regularly." | Review the benchmark route in step 5. | The benchmark scope is reviewed and observation has started. |
For a first pass, take one recommendation through to its next action. Expand the selection when you understand what the first action will help you decide.
Before you begin
Open the Context for the brand, product scope, and market you want to work on. Have one business objective in mind, such as improving consideration among small businesses or understanding your position in a new product category.
You do not need to create a new question list. Start with what SpyderBot has already generated.
If you only need to adjust the questions for a Brand Insights report, review that report's Prompt Set directly. Use this playbook when you need to decide where to spend investigation or monitoring effort across customer needs.
1. Confirm the business question and Context
Write the decision you want to make in one sentence:
"We want to understand where our invoicing product is being overlooked by small businesses evaluating software."
Make it actionable by adding the decision the evidence will inform: "We will use the results to choose which integration information to improve first." This helps you set aside interesting questions that will not change your next action.
Check that the selected Context matches the relevant products, audiences, markets, and language. A broad company Context may include needs unrelated to this decision. A Context for another product can produce reasonable recommendations that are still wrong for your current task.
If the scope is wrong, select the appropriate Context before choosing questions. If you need a distinct measurement scope, use a separate Context. Treat a scope change as a change in what you measure, not simply a new way of viewing the same history.
2. Start with SpyderBot's recommendations
In Prompt Universe, open Overview and review Next Best Actions. Then review the Recommended Analysis Set and the reasons associated with the suggested questions.
Look for three things:
- The customer need the recommendation represents.
- Why that need deserves attention.
- Whether the next step is gathering evidence, addressing an observed opportunity, or monitoring.
Prompt Priority indicates how much attention a question deserves. It does not tell you that the question has high AI search volume, that your brand is losing, or that working on it will produce a particular return.
The Recommended Analysis Set also considers coverage and redundancy. A high-priority question may already be represented by another recommended question. Do not add it back solely to make the selection match a ranking.
If several recommendations fit: keep SpyderBot's proposed order as your starting point. Prefer work tied to the decision you just defined and that your team can use now. If capacity is limited, select a coherent customer need from the recommended coverage rather than filling the budget with near-identical wording. Record any business reason for changing the order; you do not need to calculate another score.
If no recommendation fits: check the Context and available evidence first. Correct a missing business requirement or explore a relevant unobserved need. An empty recommendation list does not mean there is no opportunity.
3. Check coverage against your objective
Use Intent Landscape to see where the selection sits across customer needs and intents. Open a relevant cell to inspect the corresponding questions in Universe.
For a consideration objective, evaluation, comparison, selection, and validation questions may be especially relevant. Learning or discovery questions can still matter when they shape the customer's shortlist. You do not need a question in every cell.
Ask:
- Does the set cover the decisions our customers actually face?
- Are important constraints represented, such as business size, integration needs, or market?
- Are several questions asking essentially the same thing?
- Is a large part of the selection outside our current business scope?
Keep a recommendation when it fits. Adjust the selection only when you can explain what business need is missing or misrepresented.
4. Make only the changes that improve the decision
If the suggested set already fits your objective, continue without editing it.
If you need to change it, distinguish wording from a different customer need:
| What you notice | What to do |
|---|---|
| Equivalent questions repeat the same need and constraints. | Keep sufficient representation without adding redundant work. |
| A question refers to the wrong audience or market. | Correct the scope or adjust the question in the relevant selection. |
| An important customer requirement is absent. | Add a question that states that requirement clearly. |
| A question concerns a product you do not offer. | Exclude it from this task and record why. |
| A question shows an uncomfortable result. | Keep it if it remains relevant. Poor performance is not a reason to remove it. |
For example, "invoicing software for small businesses" and "small-business invoicing tools" may represent the same need. "Invoicing software for small businesses with multi-currency billing" introduces a requirement that could change the recommended products. Treat that distinction as meaningful.
For organic discovery, use questions that do not name your brand. A question that supplies your brand's name measures a different situation from a question in which AI must introduce it independently.
5. Match the next action to the evidence
Use SpyderBot's suggested action, then check whether the evidence supports that type of work.
These actions can form a sequence. You can investigate a question now and monitor it afterward, or improve an observed problem while tracking whether it persists. Choose the immediate next step here; you do not have to assign each question one permanent category.
| Evidence and business fit | Next step | What you want to learn or achieve |
|---|---|---|
| Important need, not explored | Explore the selected question or group. | Establish what AI currently says and which brands it recommends. |
| Important need, limited evidence | Investigate the missing scope. | Resolve a specific uncertainty, such as whether a gap exists in another AI system. |
| Observed opportunity with relevant evidence | Open the opportunity and review the proposed improvement. | Identify a justified change to content, information, sources, or the product experience. |
| A question whose changes would affect a decision | Observe it in a compatible monitoring program. | Detect meaningful changes over time. |
| Weak business relevance or redundant coverage | Defer it. | Preserve attention for the current objective. |
Do not rerun a broad analysis simply because one segment lacks evidence. Investigate the missing segment when the product supports that scope.
Before spending more credits, finish this sentence: "Another observation would help us decide whether ___." If you cannot identify a decision it could change, use the existing evidence or defer the extra run. This is a planning check, not a substitute for SpyderBot's evidence requirements.
For monitoring, define the decision that a change would trigger. "We will review our integration information if AI repeatedly describes it incorrectly" is more useful than "We want another chart."
If your immediate objective is a recurring competitive baseline, use Build AI Visibility Benchmark in Universe, review its proposed scope, and start observing from that workflow. An Explorer run is not a prerequisite.
6. Confirm the work before starting it
Review the questions, Context, AI systems, and any displayed usage or credit estimate before you start. For recurring observation, also review the monitoring scope and schedule offered by the product.
Use the existing handoff from Universe where available. Check the destination selection instead of manually re-entering the questions.
If you change the question set used for an ongoing comparison, record the change. A broader or narrower set can change the result even when AI behavior has not changed.
Example: choose the next step for an invoicing brand
Illustrative example. A team wants to improve consideration among small businesses. SpyderBot's recommendations surface the following needs:
| Customer need | Evidence available | Decision | Next checkpoint |
|---|---|---|---|
| Compare invoicing products with multi-currency support | Not yet explored | Investigate | Review which products AI recommends and the reasons it gives. |
| Validate an accounting integration | Observed responses describe the integration incorrectly | Review the opportunity | Confirm the actual capability and whether public information is clear. |
| Choose an invoicing product for freelancers | Relevant existing observations; changes could affect positioning | Monitor | Investigate a sustained change in recommendation or portrayal. |
| Find enterprise procurement software | Outside the product's current scope | Defer | Revisit only if the product scope changes. |
The team leaves the rest of the suitable recommended set unchanged. It does not rebuild the Universe or investigate every wording variation.
Record your decision
Keep the existing recommendation or result reference. Add a short decision note in your team's planning tool only when you need to coordinate work; there is no need to copy every prompt into a spreadsheet. Use this optional template for that handoff.
| Need or question group | Why it matters now | Evidence or recommendation reference | Next action | Owner and review trigger |
|---|---|---|---|---|
| Fill in | Fill in | Link to the relevant result | Investigate / improve / monitor / defer | Fill in |
You are done when each selected group has a clear next action and you can explain why it belongs in the current scope. An unchanged recommended set is a valid outcome. So is deferring a question that does not support the current objective.
Continue with: Find where competitors are preferred over your brand, if the selected work concerns a competitive gap; or Monitor the questions and changes that matter, if the next step is recurring observation. For a scope change, use Change your measurement scope without misreading progress.