Why Your Perplexity Score Is Measuring the Wrong Location

There is a failure mode in post-migration Perplexity citation data that is harder to catch than a zero. The tool runs. A score comes back. The number looks plausible. But the query ran without city context -- and for a local business, that means the score is measuring something other than local citation performance.

This is distinct from the hard failures we tracked through September 27, when the Sonar API retired and tools that hadn't migrated started returning either 403 errors or empty citation arrays. Those were visible. This one is not.

What Changed in the Agent API

When Perplexity retired the Sonar Chat Completions API on September 27, 2026, tools had to migrate to the new Agent API. The migration involved more than swapping an endpoint URL.

In the Sonar API, most AEO monitoring tools passed location as natural language inside the query prompt -- something like "You are checking local business citations in Denver for a plumber, answer as if you are a Denver resident." That text rode inside the prompt body, and Perplexity treated it as context.

In the new Agent API, location is a structured field on the search tool configuration object: `user_location`, accepting `{city, region, country}`. This field and the prompt body are separate. If a tool migrated the API endpoint but left location in the prompt text, Perplexity now processes the query without geographic grounding.

Our research team documented this in our ongoing platform investigation on 2026-10-03 (session 157), filing it as a distinct failure vector from the parsing and endpoint failures tracked through September 27: "tools that injected city/category context via prompt text may now return de-localized Perplexity results without an error signal."

No error fires. The tool reports a number. It is just not the number you think you are measuring.

Why This Shows Up as a Plausible Score

The reason this failure mode is hard to spot is that de-localized Perplexity results are not blank. They are real Perplexity citations -- just for a different competitive context.

Perplexity without location grounding answers the category query: "best plumber" instead of "best plumber in Denver." For most local service businesses, those are different answer sets. A business might appear in 3 out of 10 local Denver queries and 0 out of 10 national plumber queries. Both are legitimate Perplexity responses. But only one reflects the business's actual local citation position.

The monitoring tool's Perplexity score drops. The business owner sees a low number. The tool offers no flag distinguishing "you have no local citations" from "we asked Perplexity the wrong question."

What Perplexity Actually Returns Without City Context

There is an important mechanic here that makes the score look worse than the reality.

Our knowledge base documents a consistent finding from four April 2026 client audits: every business in that set produced a 0.0/10 Category Authority score. The explanation was the same in each case -- none of the clients had any Reddit presence. Reddit recommendation threads are the dominant citation source for Perplexity on non-localized category queries. Local directory presence, which all four clients had, does not drive citation rates when the query has no geographic constraint.

For local service businesses, the citation sources that matter -- Yelp, Thumbtack, Google Business Profile data, regional review aggregators -- are weighted toward localized queries. When location drops out, those sources matter less. Reddit, Wikipedia, national review aggregators, and domain-authority sources fill the gap.

This means a de-localized check for a local plumber or dentist will systematically under-report their actual local citation performance and over-weight their absence from Reddit. The score is coherent. It is just measuring national category authority rather than local presence.

The August Platform Shift Compounds This

There is a second layer here worth knowing.

Our methodology rec filed 2026-08-30, drawing on Pierview data covering 44,488 Reddit citations between July 1 and August 19, 2026, documented what happened after ChatGPT's retrieval pipeline reduced Reddit content dramatically on August 14. Perplexity absorbed the shift. As of that dataset, Perplexity holds 71% of cross-platform Reddit citations. A spot test on August 23 confirmed 47 out of 100 Perplexity responses cited Reddit.

The current Perplexity citation environment, for non-localized queries, is more Reddit-weighted than it has ever been. A tool running de-localized checks in October 2026 is querying a platform where Reddit content dominates the citation pool -- and most local SMBs have no Reddit footprint at all.

A local dentist in a mid-size market might have solid Yelp presence, a well-maintained GBP, and several local directories. That infrastructure performs well on local Perplexity queries. It performs poorly on national ones. A de-localized tool check currently returns a Perplexity score dominated by the very citation signal (Reddit) that local service businesses structurally lack.

How to Detect the Problem

The detection method is manual but simple.

Go to perplexity.ai directly and run two searches: one with your city name ("best [category] in [city]") and one without ("best [category]"). Check whether your business appears in both, one, or neither.

Then compare what your monitoring tool reports for Perplexity. If the city-scoped manual check shows citations your tool's score does not reflect, the gap suggests the tool may be running de-localized queries.

Ask your tool provider: does your Perplexity integration pass location via the `user_location` field in the Agent API search tool configuration, or via natural language in the prompt? Migration that updated the endpoint URL but left location in the prompt text is incomplete for local business citation tracking.

What to Do With a Score You Can't Verify

If you cannot confirm how a tool is passing location context to the Agent API, the conservative approach is to weight that tool's Perplexity data accordingly -- treat it as a national category authority signal, not a local citation signal.

The manual spot check on perplexity.ai with your city in the query remains the most reliable baseline. Perplexity.ai itself is unaffected by the API migration. It runs correctly regardless of what any monitoring platform is doing.

Sourcepull's Signal Check queries the live Perplexity interface with explicit local context -- city, category, and the query forms that match how people actually search for local services. The de-localization failure mode affects API-based tools, not browser-level checks. If you want to see how your business actually performs in Perplexity's local citation pool, that is the starting point.

The Sonar retirement created more failure modes than the obvious ones. Hard errors surface. De-localized checks do not -- they just return a score for a market you are not in.