# Analyze Amazon SQP with ChatGPT or Claude

Canonical page: https://www.kuudo.com/guides/amazon-sqp-chatgpt-claude/

Analyze Amazon Search Query Performance with ChatGPT or Claude: supply a report, compare ASIN search shares, and decide what to investigate.

Author: Kuudo

## Summary

- Download Search Query Performance (SQP) data from Brand Analytics and attach a supported file to ChatGPT or Claude, or use an authorized Selling Partner MCP connection to request the API report.
- Keep the marketplace, reporting period, ASIN, search query, counts and shares together. Ask the client to show missing fields and thin evidence before ranking opportunities.
- A difference between impression, click and purchase shares points to questions to investigate; it does not prove why shoppers behaved as they did or which listing edit will work.
- SQP does not split sponsored from organic activity or contain ad spend and ad-attributed sales. Review an advertising report separately before making paid-media decisions.

Amazon Search Query Performance (SQP) can be analyzed with ChatGPT or Claude by supplying a report or connecting authorized account data, then asking the client to compare each ASIN's search-query counts and shares. The useful result is a prioritized list of queries to investigate, with the evidence and uncertainty beside each proposed action.

The same query can look promising in one marketplace and weak in another, or appear decisive until you see it had very few clicks. Your ASIN, reporting period and team's product context determine which finding deserves attention. The [Selling Partner MCP](/features/amazon-selling-partner-mcp/) can request eligible account data, while the optional [SQP ASIN N-gram Skill](/features/skills/amazon-seller/sqp-asin-ngram/) applies a repeatable method to the ASIN and query evidence. Both paths still need your review before an action is taken.

## A scoped report gives the client something it can verify

The SQP dashboard reports how shoppers searched and how a brand or ASIN performed across impressions, clicks, cart adds and purchases. Choose the ASIN view for a product-level analysis and record the marketplace and reporting period. Brand-level and ASIN-level results answer different questions; do not mix their rows as if they represented the same product.

There are two ways to bring data into your [AI client](/features/ai-clients/):

1. **Supply a download.** In Seller Central, open Brand Analytics and the Search Query Performance dashboard. Select the ASIN scope and period, then use **Generate Download** and retrieve the report from Download Manager. Amazon says the Simple download uses the selected filters and visible columns; check that the fields you need are included. Amazon describes the download options in its [Brand Analytics announcement](https://sellercentral.amazon.com/seller-forums/discussions/t/844c405b6bf3514f57452580f5db49f2). Attach a format your client accepts; if you convert the file, preserve the column names and values.
2. **Connect an authorized account.** The Selling Partner MCP can request the Search Query Performance report for eligible sellers. Follow the [ChatGPT connection guide](/docs/quick-start/how-to-connect-amazon-seller-central-to-chatgpt/) or the [Claude connection guide](/docs/quick-start/how-to-connect-amazon-seller-central-to-claude/), then ask for the ASIN, marketplace and period you want to inspect. Amazon requires Brand Registry and the Brand Analytics Selling Partner API role for this report. The API returns one requested week, month or quarter at a time.

ChatGPT and Claude both accept CSV files, subject to the client, plan and workspace settings. Check your organization's rules for sharing account data before uploading. The upload path does not require the MCP; the connected path does not mean a report has already been fetched.

## A useful prompt keeps every finding tied to the data

The first question should establish what the client actually received. A dashboard export and an API response may use different column names or scopes. Ask the client to report the marketplace, ASINs, period, row count, available fields and missing fields before it evaluates opportunities. If the client cannot find a field, supply it or leave the finding unassessed; do not let it guess.

Copy this prompt into ChatGPT or Claude after attaching the report, or use it with an authorized connected client after requesting the report. Replace the bracketed details with your own:

> Analyze the Search Query Performance report for [ASIN] in
> [marketplace] for [reporting period]. Use the attached report or the
> report I authorized you to request.
>
> First show the report scope, row count, column names and any missing
> fields. Keep each search query tied to its ASIN and period. Check whether
> shares are fractions or percentages, and do not sum query-wide totals
> across ASINs. Ask me for anything required that is absent.
>
> Ask for my business priorities. If I have not provided them, explain
> the evidence used to order candidates and mark rows that cannot be
> ranked reliably. Do not invent a scoring rule or minimum threshold.
>
> For each search query worth investigating, show
> query volume and the available ASIN impression, click, cart-add and
> purchase counts and shares. Flag small counts or missing data as limited
> evidence; use my minimum-data rules only if I provide them.
>
> For each finding, distinguish the observed numbers from a hypothesis.
> Suggest the next evidence to check and an owner for the investigation.
> Return an action list with query, ASIN, period, evidence, uncertainty,
> next check and status. Do not infer paid ad performance, invent missing
> values, claim a cause is proven, or make account changes.

For large files, ask the client how many rows it analyzed and whether it skipped anything. Check a few output rows against the original report, including the meaning of each share. The API schema lists both query-wide totals and ASIN-specific counts and shares; those totals are not extra ASIN sales to add together. Amazon publishes the [SQP report schema](https://github.com/amzn/selling-partner-api-models/blob/main/schemas/reports/sellingPartnerSearchQueryPerformanceReport.json) for checking the connected report's fields.

## Share differences suggest where to investigate

An SQP share compares the ASIN's count with the corresponding query-wide count for the same stage and period. A change between those shares can reveal where the ASIN's share of activity differs across the search funnel. It does not identify a cause, and the shares are not the percentage of the same shoppers moving from one stage to the next.

The following is an **illustrative example**, not a customer's report. The figures show one ASIN and query for one period:

| Query | ASIN impression count and share | ASIN click count and share | ASIN purchase count and share | First investigation |
| --- | --- | --- | --- | --- |
| Example search | 600; 12% | 30; 6% | 8; 4% | Review the search-result presentation, price and delivery promise before choosing a test. |

An example action-list row for the same imaginary query makes the handoff explicit. The query-wide stage totals implied by the counts and shares are 5,000 impressions, 500 clicks and 200 purchases. Search query volume is **not provided**; it cannot be derived from those totals.

| Scope and evidence | Uncertainty | Next check and owner | Status and follow-up |
| --- | --- | --- | --- |
| Example ASIN and query, one marketplace and period; volume unavailable; impression share 12%, click share 6%, purchase share 4% | The report does not identify why shares differ; eight purchases limit confidence | Catalog analyst reviews result image, price and delivery data | Open; recheck a later comparable period |

In the example, click share is lower than impression share. That makes the result presentation and offer worth reviewing; it does not prove that the main image, title or price caused the gap. Purchase share also trails click share, so the team can examine detail-page and offer evidence after checking the actual counts. Eight purchases are a reason to be cautious about a strong conclusion, not an automatic pass or fail against an invented threshold.

The same principle applies at each stage:

- **Low impression share:** check whether the query fits the product, whether the offer is active, and whether the listing has eligibility or search issues. Advertising coverage may also be relevant.
- **Click share below impression share:** inspect the result a shopper sees, including image, title, price and delivery promise. The report alone cannot tell you which element mattered.
- **Purchase share below click share:** review the detail page, offer, price, availability and delivery. Those are hypotheses until other evidence supports them.

If the listing may be suppressed or unsearchable, check Amazon's **Fix Your Products** and search troubleshooting guidance before rewriting it. The [SQP report guide](/guides/amazon-ads-sqp-ppc-keyword-signals/) explains how this report differs from Amazon's similarly named search-term reports.

## A reviewed action list separates a finding from a change

An SQP action list is useful when another person can retrace why a query was selected. Ask the client for one row per ASIN and query with the marketplace, reporting period, observed counts and shares, evidence limitation, proposed investigation, owner, status and next review date. Keep an observation such as “click share is below impression share” separate from a hypothesis such as “the image may need review.”

Sort the list by the team's priorities and available evidence. A high-volume query with little ASIN data may need another period; a lower-volume query may matter to a strategic product. Supply any thresholds your team uses, and let the client flag rows it cannot evaluate. The optional [Search Query Analyzer](/docs/amazon-search-query-analyzer/) documents the product-specific workflow for recurring ASIN and keyword analysis.

SQP contains no ad spend or ad-attributed sales and does not split sponsored from organic activity. Do not calculate advertising cost of sales or return on ad spend from it, and do not set a bid or negative keyword on SQP alone. Bring the relevant advertising report into a separate review if the proposed action concerns a campaign.

## What happens next: review one ASIN and choose a follow-up

Start with one ASIN, one marketplace and one complete reporting period. Check the client’s field inventory, run the prompt, and verify a few results against the source. Assign each promising query to a person who can inspect the listing, offer or campaign evidence. Approve any account change through your team's normal process, then compare a later, equivalent reporting period without assuming the change caused every movement.

The goal is a list of questions your team can answer, with enough context to act on the answers.

*Next, use the [Amazon Search Query Performance report guide](/guides/amazon-ads-sqp-ppc-keyword-signals/) to keep search-funnel and paid-search evidence separate.*


## Frequently asked questions

### Do I need an MCP connection to analyze an SQP report?

No. If your account can download the report and your AI client accepts the file, you can supply the report directly. A Selling Partner MCP connection is a separate way to request eligible account data.

### Can SQP tell me which purchases came from ads?

No. Search Query Performance contains search-funnel counts and shares, but does not separate sponsored and organic activity or provide ad spend and attributed ad sales. Use the relevant advertising report for paid-media evidence.

### Does an impression-share gap identify the fix?

No. It identifies a query worth investigating. Relevance, listing eligibility, stock, offers and advertising are possible lines of inquiry; compare the actual account and product evidence before choosing an action.


## Sources

- [Amazon: Download Brand Analytics search dashboard data](https://sellercentral.amazon.com/seller-forums/discussions/t/844c405b6bf3514f57452580f5db49f2)

- [Amazon: Analytics report type values](https://developer-docs.amazon.com/sp-api/docs/report-type-values-analytics)

- [Amazon: Search Query Performance report schema](https://github.com/amzn/selling-partner-api-models/blob/main/schemas/reports/sellingPartnerSearchQueryPerformanceReport.json)

- [OpenAI: Data analysis with ChatGPT](https://help.openai.com/en/articles/8437071-data-analysis-with-chatgpt)

- [Anthropic: Upload files to Claude](https://support.claude.com/en/articles/8241126-upload-files-to-claude)