SERP Tracking

Overview

SERP Tracking captures search engine result pages alongside your AI snapshots. It's how you correlate "what AI engines say about you" with "what shows up in classic search," because most AI engines pull from the same SERP sources. If a page ranks page 3 in Google, it's unlikely to be cited by Perplexity or Gemini either.

Plan gate: Native Google AI Overview and AI Mode capture needs only the serp_tracking feature, which turns on at Pro-Individual and above — AEO Optima captures these surfaces itself (in India with its own browser, elsewhere through its search partner), so no connector is required. A SERP provider connector (Serper or DataForSEO) is only needed for Bing Copilot or as an alternative source, and connectors are available from Pro-SME and above.

What gets captured

For each tracked keyword, on each scheduled run, the platform captures:

Configuration

Open Settings → SERP Tracking to configure capture.

Provider

A SERP provider is optional — AI Overview / AI Mode capture is built in and needs no connector. Connect a provider only for Bing Copilot, or if you already subscribe to one and prefer it as an alternative source. Supported:

• DataForSEO — returns the full AI Overview citation list, has a dedicated AI Mode endpoint, and covers Bing organic SERP; granular country/language coverage. AI Overview / AI Mode use the Advanced endpoint (~2× the standard per-query cost).
• Serper — fast, flat per-query pricing, US/EU regions. Best for organic SERP; AI Overview is best-effort only (answer-box fallback) and there is no AI Mode.

When a provider is connected it takes precedence over the built-in capture for AI Overview / AI Mode (your account, your billing), and the pipeline prefers DataForSEO for those surfaces and uses either for plain organic.

Add credentials at Settings → Connectors. Without credentials configured, the built-in AI Overview / AI Mode capture still runs — only Bing Copilot and connector-only sources are skipped.

How AI Overview and AI Mode are captured

• In India, AEO Optima captures Google AI Overview and AI Mode with its own browser, through an Indian residential IP. From 22 October 2026, an India question Google blocks on that route (its anti-bot page) is answered through our search partner instead, in the same run, and marked so: the answer page shows the search partner as how it was captured, and the run says how many questions Google blocked and how many our search partner answered. A failure in our own capture system is not sent to our search partner; the run says so and our team is alerted.
• In every other market, from 7 October 2026, your scheduled run captures them through our search partner, DataForSEO, which asks Google from that market itself (DataForSEO's live endpoint). Our own browser is not used outside India: Google does not give it real answers there.
• Each answer shows how it was captured ("How this answer was captured" on the answer page), and each run lists how many search answers came from each route. A market our search partner cannot be asked for is named on the run as not captured, never asked as another country. Our search partner serves AI Overview and AI Mode only: Bing Copilot still needs a connection of your own, and without one the run names it as not captured. From 8 October 2026, if the search step cannot be prepared for a run, the run says so ("the search step could not be prepared") rather than showing no search answers.
• If you connect your own DataForSEO or Serper account, it is used instead, in every market.
• Your organisation can ask us not to send its questions to our search partner. We then never do: outside India your runs capture AI Overview and AI Mode only through a search connection of your own, and without one each run says "no search source for this market". In India, a question Google blocks on our own route is then not captured that run.

What counts as a mention on Google's AI surfaces

On Google AI Overview and AI Mode, only the AI answer's own text counts. Your brand is named, and a competitor is counted, only when that text names them. The tone of an answer, its fact check and the brands it names are all judged on that text alone.

• The sources an AI Overview cites are citations, not mentions. A brand that appears only in the title of a cited page is not named in the answer. Being cited is shown separately, as citation presence.
• The ordinary search results below the AI answer are not the AI answer. A brand that appears only there is not named. We record it as organic presence, which is never counted as a mention.
• When Google shows no AI Overview, there is no AI answer. No brand is named, no competitor is counted, and nothing is judged.

From 9 October 2026, organic results and an AI Overview's source list no longer count. Before that date, a brand or competitor that appeared only in the organic results under an AI Overview was counted as named in it, and the tone, fact-check and brand judgements read those results too.

Locations

Add the geographic locations you want to track. Each location has:

• Country code (e.g., US, GB, DE)
• City (optional, for local SERPs)
• Language (e.g., en, es, fr)

Most providers charge per query and per location separately, so adding more locations multiplies your SERP capture cost. See Locations.

Devices

Toggle desktop, mobile, or both. Mobile SERPs often differ significantly from desktop and tend to be what AI engines actually index, so we recommend tracking mobile by default.

Keyword sets

A keyword set is a list of queries to track. You can:

• Add keywords manually (one per line)
• Import from a CSV
• Auto-generate from your prompts via the Query Universe export

Each keyword in each (location × device) combination uses one query credit per cycle.

Schedule

SERP captures run on the same schedule as your snapshots — see Schedules. You can set them to daily, weekly, or custom cadence per project.

The Google AI surfaces follow that same schedule. In India, AI Overview and AI Mode are captured with our own browser rather than through a connector, and that capture is retried shortly after your scheduled run to recover any surface the first attempt could not reach — a Google block, or a residential IP that could not be trusted. The retry is part of your scheduled capture, not a second one: it runs only on the days your schedule runs, it skips anything already captured that day, and a project whose schedules are all switched off is not captured at all.

So a weekly project is captured once a week on every surface, native ones included. If you pause a project, it stays paused.

To run a one-off capture, use the Capture now button on the SERP Tracking page or trigger it via the API.

Cost considerations

SERP capture has a cost per query that varies by provider. The math for budgeting:

queries_per_run = keywords × locations × devices
runs_per_month = depends on schedule (daily = ~30, weekly = ~4)
monthly_cost = queries_per_run × runs_per_month × per_query_cost

The platform tracks SERP costs in Cost Tracking under the serp source type. Watch this number — it's the easiest line item to accidentally blow up by adding "just one more location."

Best practices

• Start narrow. Pick the 20-30 keywords that matter most before scaling up. A small, well-tuned set is more useful than a sprawling one you don't review.
• Track mobile, not desktop. Mobile SERPs are closer to what AI engines see, especially for Google.
• Don't duplicate locations unnecessarily. Tracking US and New York, US is rarely worth the extra cost — pick one.
• Use Query Universe exports for keyword sets. The Query Universe knows what your prompts are about; let it pick the keywords for you instead of guessing.
• Audit costs weekly. SERP costs can grow silently if someone adds a new location or doubles the keyword count.

Related docs

• Schedules — when SERP captures run
• Locations — geographic targeting
• Connectors — adding Serper / DataForSEO credentials
• Cost Tracking — monitoring SERP spend