SerpApi released AppTrail on September 30, 2026, an open-source program that records where a mobile app ranks in the App Store and Google Play and whether AI-generated answers mention it, according to a post on the company's blog.

In Short

SerpApi, a company that sells search-result data, released a program that checks where a phone app shows up in app-store searches and in AI answers, then saves every result. It matters to app makers and marketers because rankings and AI mentions move often, and a saved history shows what changed and when. The program runs on a personal computer or a server, and each check spends credits from a SerpApi account, with 250 free searches a month on the free plan, according to SerpApi.

What was released, and by whom

Adarsh Divakaran, identified in the post as SerpApi's Python Developer Advocate, wrote the blog entry, which carries a September 30, 2026 date. According to SerpApi, the program is distributed through PyPI and Docker Hub, with the source code on GitHub under an MIT license. The repository describes the project as "Open-source app visibility tracking across the App Store, Google Play, and AI search." At the time the page was captured it listed release v1.0.0 as the latest of three releases, 12 commits, two branches, three tags, eight stars, no forks and one contributor, Divakaran himself. The language split reads Python 58.7%, JavaScript 29.3%, CSS 10.4%, HTML 1.5% and Dockerfile 0.1%.

The commercial arrangement is part of the story. The code is open-source, yet every data point it stores arrives through SerpApi's metered API, which returns structured SERP data. According to the README, the program fetches results through the operator's own SerpApi account, and discovery, product verification and tracking requests can all consume credits. The publisher of the code is also the vendor of the data feed on which it depends.

What the program records

Store rankings and competitor comparisons

Keyword ranking is the core function. A user connects a SerpApi key, picks an iOS or Android app by name or store URL, adds search terms such as "habit tracker", and the program logs the app's position in the App Store and Google Play. An Overview chart plots average store position over time, and any individual check can be opened to read the saved results behind a figure. A Distribution view sorts outcomes into positions 1-3, 4-10, 11-50, 51 and beyond, and not found.

The change counts follow a defined rule. According to the README, the latest check for each app and query combination is compared with the preceding period of equal length, and only combinations checked successfully in both periods contribute. Failed checks and featured-only placements appear separately.

Competitors are added beneath any search. The blog demonstrates the feature with an invented app, HabitFern, tracked against two invented rivals, RoutineCove and StreakPebble; in that example HabitFern sits ahead of both for "habit tracker" in the US App Store. A competing app that already exists in a workspace reuses saved results and shares future checks, so it adds no tracking searches, according to the README. Discovering a new app, verifying its listings and refreshing store details by hand can use separate credits. Store comparisons need a verified listing on the matching platform.

Countries, languages and devices

A query matrix places each country's saved results beside those of the competitors, with a check time in every cell and a link to the underlying evidence. Results for different languages, devices and search depths stay apart. Bing Copilot returns global results rather than country-level ones, and appears in the Sources view.

AI mentions and citations

For questions such as "Which apps help build daily habits?", the program queries Google AI Mode, Google AI Overviewsand Bing Copilot. It stores the answer, the matching text and the cited sources. The blog separates two outcomes: a mention means the app appeared in the answer, while a citation means the answer also linked to the app's store listing or website. The documents name no other AI sources.

Listing history

Collection of listing changes is off by default, in the README's words ("Collection is off by default."). A user selects an app, store, country and frequency, with a language option for Google Play, and a credit estimate appears before tracking starts. Each scheduled check uses one SerpApi product request, plus an initial baseline check; manual checks and retries can consume more.

The fields followed include titles, descriptions, versions, pricing, icons and screenshots. Text comparisons highlight additions and removals, and screenshots are compared in order, with added and moved images labeled. Supported images are archived in the local database. On missing history, the README is categorical: "It never substitutes a live image for a missing historical image." Collection begins only when enabled, so earlier listing versions cannot be reconstructed.

Alerts, schedules and exports

The notification bell reports a drop of at least five positions from a previous top-10 rank, an exit from the top 10, or disappearance from a sufficiently deep set of checked results. "Two successful checks must confirm a loss." A failed check in between resets the confirmation, and continuing losses do not generate duplicate alerts until the ranking recovers. Alerts, matrices and distributions read saved data and use no extra SerpApi credits, according to the README.

Checks can be run on demand or scheduled daily, weekly, every two weeks or monthly. Results export as CSV from Settings, and workspace backups can be downloaded; a restore option from SQLite backups appears in the repository's commit history.

Installation and technical requirements

Python route

AppTrail needs Python 3.11 or newer and a SerpApi API key, entered in a setup wizard. It runs through uvx apptrail, or after pip install apptrail and the command apptrail. Either way, the web interface opens on a free localhost port. On the first run a setup code printed in the terminal creates the owner account. Account and database sit in a shared user data directory, which survives restarts and package upgrades.

Docker route

The Docker image is serpapi/apptrail:latest. The README's command runs it detached under the name apptrail, maps port 80, mounts a named volume, apptrail-data, at /data, restarts the container unless it is stopped, and sets a 360-second stop timeout. The one-time setup code appears in the container logs. According to the README, the image supports Intel/AMD and ARM Linux; if port 80 is taken, the mapping changes to 8080:80. A separate self-hosting guide covers CapRover, Coolify, Docker Compose, building from source, HTTPS, backups and updates, and the blog links a third-party Coolify 4 walkthrough for virtual private servers.

SerpApi describes the Python install as suited to local testing and quick checks, and Docker on a machine that stays running as the route for continuous scheduled monitoring. The process or container must keep running for scheduled checks, though the browser can be closed. If it stops, saved history remains and overdue checks resume on restart. The README adds a limit: "Missed historical results cannot be reconstructed."

Access control

The dashboard and every workspace API, including searches, exports and backups, require login. Registration closes once the owner account exists, new server runs need a fresh login, and a forgotten password is recovered from the server terminal with apptrail --reset-password.

Credits and costs

According to SerpApi, its free plan includes 250 search credits per month. The blog calculates that weekly tracking of an app and competitors, for two keywords in both stores across three countries, plus two questions across the three AI sources, would use up to 160 credits over five weekly runs at default search depth. That is a vendor-supplied estimate, and the documents do not show the arithmetic. The README says Settings displays a monthly usage estimate and the account-wide remaining balance, and the setup flow shows a usage estimate before tracking starts. Identical tracked searches are shared across apps. Pricing beyond the free plan is not covered in either document.

Where the two documents differ

Two inconsistencies appear. The blog states that the backup download saves the workspace "including its history and archived listing images." The README says downloaded backups omit archived images and raw SerpApi responses to save space, while keeping saved results, matched evidence, listing text and image-change records, and that after a restore the program explains why those items are unavailable. The documents do not reconcile the two descriptions.

The second concerns the Docker command. The blog's version has four flags; the README's adds --stop-timeout 360. The effect of omitting the flag is not described.

The release arrives while SerpApi is a defendant over how it collects Google results. Google sued SerpApi on December 19, 2025, and on July 20, 2026 a federal judge dismissed every claim in the original complaint, letting Google refile a narrowed claim. Google then filed an amended complaint on August 10 built on licensing agreements, and SerpApi asked on August 24 for the case to end with prejudice, with a hearing set for September 29, 2026. Later PPC Land reference material lists October 13 for that hearing, so the current date is not confirmed here. Separately, Reddit sued SerpApi, Oxylabs, AWMProxy and Perplexity AI on October 22, 2025, and SerpApi filed antitrust counterclaims on August 28, 2026. Neither attached document mentions any of this.

The technical side has tightened too. Google removed the num=100 parameter on September 14, 2025, turning one request for 100 results into ten, and restricted SerpApi's Light Fast API to three results in late September 2025. Google then confirmed on August 26, 2026 that it routes search result links through google.com/goto addresses, which Nozzle measured at 500 to 1,000 requests to resolve a five-page ranking. SerpApi said on September 5 that it had deployed a fixrestoring direct destination addresses. The attached documents do not say which SerpApi endpoints feed each AppTrail source, so any effect of the redirect on AppTrail is not stated.

Why it matters for marketers

Measurement of AI visibility remains unsettled. The IAB's August 3, 2026 framework, as PPC Land covered it, found that only 16% of brands systematically track AI visibility, with more than 20 vendors selling tools whose methods return different answers for the same brand. A Semrush survey found 45% of respondents unable to measure their brand's visibility in AI-generated answers properly, and only 9% able to measure all the relevant metrics. Microsoft Clarity brought Citations to general availability on May 13, 2026 and added Topic Insights on July 9, free to all users, but those tools cover websites. AppTrail's remit is store listings plus three AI sources, a narrower and different surface. That Google's AI Mode has reached one billion users helps explain why an app-ranking tool now includes it.

Variability is the harder problem. The IAB document illustrates that a brand's share of voice can move five points overnight because a model was retrained, and that a four-point movement means something only if it exceeds the variance from re-running identical queries on the same day. AppTrail's design choices - timestamped saved evidence, a two-check rule before alerts, exclusion of failed checks from change counts - target single-run noise. The documents do not cite the IAB framework or quantify how much the three AI sources vary between runs.

On the store side, PPC Land's coverage of Apple's June 2026 developer conference listed new App Store marketing surfaces, among them Creative Assets, Personalized Collections, App Notes and Featuring Nominations, and noted that organic and paid app teams now share more of the same canvas. Keyword position is one input among those surfaces, and the program logs position and listing text but not the marketing assets Apple described.

Several limits stand out. Tracking depends on one vendor's API and credits. History starts only when a query or listing is enabled, and gaps while the process is down cannot be filled. The AI coverage stops at Google's two surfaces and Bing Copilot. And the repository is a v1.0.0 release with one listed contributor and eight stars at capture, a maturity indicator rather than a verdict.

Timeline

  • September 14, 2025 - Google removes the num=100 parameter, multiplying requests needed for 100 results by ten
  • Late September 2025 - Google restricts SerpApi's Light Fast API to three results
  • October 22, 2025 - Reddit sues SerpApi, Oxylabs, AWMProxy and Perplexity AI
  • December 19, 2025 - Google sues SerpApi over search scraping
  • May 13, 2026 - Microsoft Clarity Citations reaches general availability
  • June 23, 2026 - first public sighting of google.com/goto rewritten result links
  • July 9, 2026 - Microsoft Clarity adds Topic Insights, free to all users
  • July 20, 2026 - judge dismisses all claims in Google's original complaint against SerpApi
  • August 3, 2026 - IAB publishes Measuring Visibility in the AI Era
  • August 10, 2026 - Google files an amended complaint against SerpApi
  • August 24, 2026 - SerpApi asks the court to end the case with prejudice
  • August 26, 2026 - Google confirms the goto link rollout
  • August 28, 2026 - SerpApi files antitrust counterclaims against Reddit
  • September 5, 2026 - SerpApi says it deployed a fix restoring direct destination URLs
  • September 29, 2026 - hearing scheduled on SerpApi's motion to dismiss Google's amended complaint
  • September 30, 2026 - SerpApi publishes its blog post releasing AppTrail, with the GitHub repository listing release v1.0.0

Summary

Who: SerpApi, the Austin, Texas search-data provider, through Adarsh Divakaran, its Python Developer Advocate, who wrote the post and is the repository's only listed contributor.

What: AppTrail, an MIT-licensed program that records App Store and Google Play keyword rankings, competitor positions, listing changes, and mentions and citations in Google AI Mode, Google AI Overviews and Bing Copilot answers, with alerts, scheduled checks and CSV export. It runs on Python 3.11 or newer or in Docker and spends credits from the operator's SerpApi account, which includes 250 free searches a month.

When: The blog post is dated September 30, 2026. The repository lists release v1.0.0 as the latest.

Where: On a personal computer or a self-hosted server, with packages on PyPI and Docker Hub and source code on GitHub.

Why: According to SerpApi, people find apps through store searches, browsing and questions put to AI tools, and a single check says little about change over time. The program saves each result with its evidence so that movement in rankings and AI mentions can be inspected afterwards.