AI Visibility Tool Migration Checklist: Switching Between Profound, Otterly, and Peec Without Data Loss

TL;DR

Switching AI visibility tools without data loss takes three things: a clean export of your prompt set and citation history before you cancel, a structured replay in the new platform with identical prompts and competitor lists, and a 30-day dual-run window so you can reconcile baselines. Skip any of these and you lose the trend story your last quarter was built on.

When you actually need to migrate

Migration is not the same as adding a tool. Migrate when:

• Pricing or seat limits no longer fit (Otterly starts around $29/month, Peec around €400/month, Profound is enterprise-tier).
• Platform coverage gaps matter — for example, you need Google AI Mode or Microsoft Copilot tracking the current vendor does not support.
• Methodology changed and you no longer trust the data (UI-scrape vs API sampling produces different citation counts).
• Your team needs features the current platform does not ship: bulk prompt analysis, MCP integration, agent automations, or stronger CSV exports.

If the trigger is just "the dashboards look prettier elsewhere," run a 14-day trial in parallel before committing to a full migration. See the AI visibility tool buyer's checklist for selection criteria you should re-validate before switching.

The migration checklist

1. Inventory what you have today

• [ ] Export the full prompt set from the source tool, including tags, country, and intent volume where available.
• [ ] Export the citations table — prompt, platform, cited URL, position, date, brand mentioned, competitors mentioned.
• [ ] Snapshot the competitor list and any custom groupings.
• [ ] Capture alert rules: drop thresholds, share-of-voice triggers, Slack/email recipients.
• [ ] Save dashboards as PDF or screenshot the executive views you actually use.
• [ ] Note the methodology metadata the tool reports (UI scrape, API sampling, sample size, refresh cadence) so you can compare apples to apples later.

Otterly.AI exposes Search Prompts and Citations CSV exports from the Brand Report. Peec AI surfaces exports through dashboards and its MCP integration. Profound supports CSV export from Prompt Volumes and Brand Relevant Prompts. Always pull the most granular export the tool offers — summaries lose evidence.

2. Decide what counts as the system of record

• [ ] Pick a canonical store for historical data: a Notion database, a BigQuery table, or a Google Sheet your team actually maintains.
• [ ] Standardise the schema: date | prompt_id | prompt_text | platform | brand_mentioned | competitor_mentioned | cited_url | position | source_tool.
• [ ] Backfill with the CSV exports from step 1 before you touch the new platform.
• [ ] Tag every row with source_tool so future reconciliation is unambiguous.

This step matters because no AI visibility vendor backfills history when you onboard. Otterly explicitly states it tracks prompt performance only from the moment you create a project — it does not import past data. The same is true for Peec and Profound. If you cancel before exporting, your trend line is gone.

3. Rebuild the prompt set in the destination tool

• [ ] Recreate prompts verbatim. Even small wording shifts change citation results across LLMs.
• [ ] Match country and language segmentation exactly.
• [ ] Re-attach tags (campaign, funnel stage, product line) so reporting maps cleanly.
• [ ] Add the same competitors in the same order; competitor framing affects share-of-voice math.
• [ ] Validate that the destination tool covers every AI platform the source tool covered. If coverage is narrower (e.g., no Copilot), document the gap before cutover.

Use a prompt-set design checklist to audit and prune stale prompts during this step — migration is the cheapest moment to clean up.

4. Run both tools in parallel for at least 30 days

• [ ] Keep the source tool live and paying for one full reporting cycle.
• [ ] Compare daily: visibility %, citation count, share-of-voice, sentiment.
• [ ] Expect 10-25% variance between platforms — different sampling and prompt timing produce different numbers even on identical prompts.
• [ ] Document a reconciliation note explaining the variance so future readers of your dashboards do not panic at the discontinuity.
• [ ] Re-run alerts in both tools and compare false-positive rates before trusting the new alert layer alone.

5. Cut over and decommission cleanly

• [ ] Pull a final export from the source tool the day before cancellation.
• [ ] Confirm the canonical store has the merged history (source + destination overlap window).
• [ ] Update runbooks, Looker/Mode dashboards, and Slack channels to point at the new tool.
• [ ] Revoke API keys, MCP tokens, and SSO entries for the retired platform.
• [ ] Schedule a 30-day post-migration review to verify alerting, citation grounding, and reporting cadence.

What does not transfer between tools

• Historical prompt-result timeseries — every tool starts its clock at project creation.
• Sentiment scores — each vendor uses a different model and scale.
• Share-of-voice formulas — Peec, Profound, and Otterly weight mentions, citations, and position differently.
• Alert state — paused alerts and snooze rules never migrate.
• Custom annotations and dashboard layouts.

Treat all of these as rebuild work, not import work. For a deeper view of how share-of-voice differs across vendors, see share of voice in LLMs.

Common migration mistakes

• Cancelling the source tool the day you sign the new contract. You lose the export window and the dual-run baseline.
• Trusting day-one numbers from the destination tool. Most platforms need 7-14 days of crawls before metrics stabilise.
• Reusing the same prompts without auditing them. Migration is the cheapest moment to retire stale prompts and add fresh ones.
• Forgetting to migrate competitor sets. Competitor changes alone can shift visibility scores by 15+ percentage points.
• Skipping the canonical store. Without an external system of record, the next migration repeats the same data-loss problem.

FAQ

Q: Can I import historical data from Profound, Otterly, or Peec into a new AI visibility tool?

No native imports exist between these platforms as of 2026. Each tool starts tracking from the moment a project is created. The portable layer is CSV exports — prompts, citations, and competitor sets — which you load into your own canonical store, not into the destination tool's database.

Q: How long should I run two AI visibility tools in parallel?

At least 30 days, ideally a full reporting cycle (often 4-6 weeks). That window lets you reconcile baselines, validate alerting, and build a variance note for stakeholders. Cutting parallel running shorter than 30 days almost always produces an unexplained "drop" in dashboards the quarter after migration.

Q: Which export format should I prioritise?

Granular CSV at the prompt + citation level. Summary exports drop the evidence trail you need if leadership later questions a visibility change. Otterly's Citations Full Report, Peec's prompt-level exports, and Profound's Prompt Volumes CSV all expose the row-level fields you need.

Q: Will migrating change my visibility score?

Usually yes — by 10-25%. Different vendors sample at different times, weight citations differently, and cover different AI platforms. Document the variance before cutover so the change is not mistaken for a real visibility drop.

Q: Do I need to keep the old tool indefinitely for compliance?

No, if you have CSV exports stored in your canonical system of record. Most teams retain exports for 24 months and decommission the source tool within 60 days of successful migration.