The ChatGPT Citation Recovery That Doesn't Apply to Most Local Queries

Across the AEO community in July 2026, there's been a wave of optimism about ChatGPT. GPT-5.6 Sol uses the site: operator on up to 71% of searches -- meaning it pulls from brand websites directly instead of routing through third-party sources. Brand citations are climbing. The narrative: schema is back.

For enterprise clients, B2B software companies, and professional services firms, this is genuinely meaningful. For a plumber in Mississauga or an HVAC company in Burlington, it changes almost nothing. The reason comes down to who's actually submitting those queries.

What the Sol reversal actually says

In our July 20-23, 2026 research cycle (Scout sessions 82 and 85), we tracked the GPT-5.6 Sol citation finding from its initial publication through follow-up confirmation.

The core data: GPT-5.5 used the site: operator on only 12.6% of searches -- a sharp drop from 40.5% in GPT-5.4 Thinking. GPT-5.6 Sol reverses that, with site: operator usage back at up to 71%. Brand citations from the business's own website are meaningfully more likely under Sol than under GPT-5.5. The third-party-first framing that shaped ChatGPT optimization guidance for most of 2026 is less accurate for Sol than it was.

That's a real finding. The question is who it applies to.

Sol is a paid-tier feature in standard conversation

Our July 23 follow-up (Scout session 85, `methodology-recs/2026-07-23-gpt56-sol-free-tier-clarification.md`) identified the scope problem with the Sol reversal finding.

GPT-5.6 Sol is not what most ChatGPT users interact with in a standard query. In plain-chat conversations, the model breakdown works like this: free and Go plan users receive GPT-5.5 Instant by default; Plus plan users receive Sol as their primary model; Pro, Business, and Enterprise users can access Sol and specialized tiers.

A homeowner typing "best HVAC company in Hamilton" into the ChatGPT app is almost certainly on the free plan. The response they receive comes from GPT-5.5 Instant, where the third-party-first methodology still applies. Schema markup and on-site structure are weak signals at that tier. Directory presence -- Yelp, HomeStars, BBB, Houzz -- is the primary citation driver.

The Sol reversal applies to a paying-subscriber segment. For home services consumer queries, our session 85 assessment is specific: Sol-tier queries come from higher-intent buyers -- business owners comparing vendors, professionals deciding on contractors. These are not the volume of residential service appointment calls that keep an HVAC company running.

The direct conclusion from session 85: "The GPT-5.5 third-party-first framing (Methodology Rec 2026-06-16) remains correct for the free-tier majority." The Sol reversal is real but does not change the fix priority for most local service businesses.

Perplexity doesn't have this tier fragmentation

Our July 24, 2026 methodology update (Scout session 86, `methodology-recs/2026-07-24-perplexity-home-services-format.md`) adds a clarifying contrast.

Perplexity doesn't have ChatGPT's tier structure. Free and paid users get the same live-retrieval model. For local services queries -- plumbing, HVAC, electrical, roofing -- the DeltaV Digital AI Citation Study (25,337 citations tracked from April through July 2026) found that homepages account for 55% of AI citations. A separate May 2026 local query study found Perplexity cites business websites directly 73% of the time for "best [service] in [city]" queries.

This is not in conflict with the ChatGPT finding. It's a different platform with different citation logic. For Perplexity, homepage optimization is the primary lever -- and it applies uniformly across user tiers. For ChatGPT free-tier (the majority of residential customer queries), directory presence is the primary lever. The Sol reversal makes brand-site schema more relevant for the paid-tier ChatGPT segment, but that's a secondary audience for most home services businesses.

What the data doesn't support: prioritizing schema markup over directory work for a residential plumber or electrician whose customers are primarily free-tier ChatGPT users. That investment belongs higher in the fix plan for the paid-buyer segment, not the general residential customer.

The check that comes before any of this

One failure mode makes all platform-specific optimization irrelevant: if AI crawlers cannot access your website, schema and content changes produce no effect on citation rates.

Our July 21, 2026 methodology update (Scout session 83, `methodology-recs/2026-07-21-cloudflare-ai-bot-blocking-preflight.md`) identified this as an unaddressed gap in standard AEO fix plans. Many businesses have GPTBot, PerplexityBot, or ClaudeBot blocked -- via a line in robots.txt, a plugin that toggled it automatically, or a CMS that defaulted to blocking AI crawlers.

Starting September 15, 2026, all new domains joining Cloudflare will have Training and Agent bots blocked by default on ad-carrying pages. The classification of Perplexity's live search crawlers under Cloudflare's taxonomy -- Search bots (which remain free) or Agent bots (which will be blocked) -- is not yet confirmed. But existing robots.txt blocking is already active for businesses that have set it manually or whose CMS set it automatically, and Sourcepull's current audit pre-flight does not check for it.

Confirm AI crawlers are allowed before any other fix runs. Schema improvements on a bot-blocked website have no effect.

The actual fix sequence for local service businesses

From our three July investigations, the priority order for a home services business serving residential customers:

First, confirm AI crawler access. Check robots.txt for GPTBot, PerplexityBot, and ClaudeBot. This is the gate. Everything else depends on it.

Second, for ChatGPT -- where the free-tier majority lives -- directory presence is the primary lever. Yelp, BBB, HomeStars, Houzz, and any category-relevant platforms. NAP consistency across all of them matters more than what's on your homepage.

Third, for Perplexity -- where tier doesn't vary -- homepage optimization is the primary lever. Direct service answers in the first 100 words, LocalBusiness schema, PerplexityBot allowed in robots.txt.

Fourth, for ChatGPT Plus and paid-tier buyers -- the segment where Sol applies -- brand-site schema and on-site structure become a higher priority. The Sol reversal is real and relevant here.

The Sol reversal is not a reason to reprioritize schema over directories for most home services clients. It's a reason to add schema earlier in the fix plan when the known buyer profile skews toward paid ChatGPT users.

Generic AEO advice flattens across platforms and tiers. The specific failure depends on which platform your customers use and which model tier they're on. Signal Check at sourcepull.ca runs live queries for your business across ChatGPT, Perplexity, Gemini, and Claude and surfaces which platform isn't citing you -- not a checklist, a platform-specific diagnosis.