01

Field research

What actually gets an AI answer to mention a Doha business

We ran a real, two-engine scan across two Doha categories, dental clinics and hotels, and logged exactly which sources ChatGPT and a web-grounded search pass cited when asked the questions a buyer actually types. Here's what showed up, and what it means for getting cited yourself.

02

The receipt

Scan date
17 July 2026
Categories
Dental clinics and hotels, Doha
Engines used
ChatGPT (gpt-4o-search-preview, live OpenAI web search) plus a Google-grounded web pass, standing in for how other AI search tools retrieve with grounded web search
Samples per question
2 runs per question
Scope
Real, partial scan, not the full product. See the caveats below.

This study used two engines. The full Awal product measures five:

  • ChatGPT
  • Claude
  • Perplexity
  • Gemini · grounded in Google Search
  • Fanar

03

The pattern: what AI actually cites, in order

When someone asks an AI assistant a "best [category] in Doha" question, here's the order the sources actually got cited in, across both scans, both engines:

  1. 01

    Google

    Business Profile, Maps, Hotels, and reviews. This was the most-cited single source in both scans by a wide margin: ChatGPT’s own answers cited google.com six times across the dental questions and twelve times across the hotel questions.

  2. 02

    Category directories and review platforms

    The listings built specifically for the vertical: tabeebak.qa, heliumdoc.com, and whatclinic.com for dental clinics; TripAdvisor, Booking.com, and tripexpert.com for hotels.

  3. 03

    Editorial “best of” listicles

    Local blogs and travel media: majddental.net and qatarstalk.com for dental; Time Out Doha, Michelin Guide, Forbes Travel Guide, karlieplace, and mel365.com for hotels.

  4. 04

    Community forums

    Qatar Living threads for dental questions, TripAdvisor forums for hotel questions. This is the source AI leans on for “which one do people recommend”. Reddit didn’t surface in this US-biased tool, which is a gap in the tool, not proof Reddit doesn’t matter here.

  5. 05

    The business’s own site

    Cited last in both scans, and mostly only for questions that already named the business: branded searches and head-to-head comparisons.

This matches what wider GEO research already predicts: AI weights earned third-party mentions (directories, review sites, editorial coverage) far above a business’s own marketing. Getting an AI answer to name a business means getting that business onto the sources AI already trusts. Your own homepage isn’t where that happens.

Getting cited means showing up on the sources AI already trusts, not writing a better homepage. In this scan, a business’s own site was the last source either engine reached for.

04

Two scans, one pattern

The two businesses below are anonymized. This page shows how AI treated them, not who they are. Every source, question, and rate is real and unedited.

Scan 1: a dental clinic in Doha

Buyer questionChatGPT (search)Web-grounded pass
best dental clinic in Doha1 / 2Absent from the top 10
best dentist for implants0 / 2Absent
top cosmetic dentist in DohaNot sampledAbsent
affordable dental clinic reviewsNot sampledAbsent
which clinic do people recommend2 / 2Absent (Qatar Living threads dominate)

The clinic in this scan is strong wherever reviews are the ranking signal. On Google and one AI-driven directory, it carries a near-perfect rating across a large, sustained volume of reviews. It’s effectively invisible wherever directories, forums, or editorial lists are the signal instead.

Where the gaps are

SourceTypeStatusWhat closes the gap
Google Business Profile / reviewsReviewsStrongMaintain review velocity; keep every branch’s listing complete.
tabeebak.qaHealth directoryBranded page only, absent from the category and treatment-tag browse pagesGet listed in the category pages. It’s the cheapest win available.
heliumdoc.comDoctor directoryAbsentEnroll the clinic and its dentists.
whatclinic.comPrice / review aggregatorAbsentClaim and list. This is what feeds “affordable” and “reviews” questions.
Qatar Living forumsCommunityNo organic mentions foundEarn genuine patient recommendations, the source “who do people recommend” answers ground in.
majddental.net / qatarstalk.comEditorial “best of”Present on one list, absent from the other’s “Best 10”PR outreach to replicate the win on the list it’s missing from.
zavis.aiAI directoryStrongKeep every branch indexed and current.
The clinic’s own siteOwnedLoses head-term searches to competitors’ own SEO pagesA separate on-site SEO pass. Worth doing, but not what wins the AI answer.

Scan 2: a five-star hotel in West Bay, Doha

Buyer questionChatGPT (search)Web-grounded pass
best 5-star hotels in Doha0 / 2Absent (other international luxury brands lead)
best luxury hotel in West Bay0 / 2Absent
this hotel vs. its three rivals, by name2 / 2 (cited tripexpert.com + TripAdvisor)Present, the question already named the hotel
top-rated hotel reviewsNot sampledPresent (Booking.com, 8.4, “excellent”)

The one question where the hotel appears reliably is the one that already named it: a head-to-head comparison against three rivals. That’s the query design doing the work, not a citation win worth celebrating on its own.

On TripAdvisor, the most-cited hotel source in this scan, the hotel ranks in the top 30 of more than 200 Doha properties, ahead of three rival hotels. Yet it’s absent from every editorial “best hotels in Doha” list, which is exactly what AI summarizes when a buyer asks that question. The fix here is editorial placement, not more reviews.

Where the gaps are

SourceTypeStatusWhat closes the gap
Google Hotels / reviewsReviewsStrong, thousands of reviews at a four-star averagePush review recency to lift the “top” modules.
TripAdvisor (+ tripexpert.com)Reviews / comparisonTop 30 of 200+ Doha hotels, beating three rivalsLead with this. Get head-to-head comparison pages built on tripexpert.
Time Out Doha, Michelin Guide, Forbes Travel Guide, karlieplace, mel365.comEditorial “best of”Absent from the major lists; present on mel365.com only, for value and nightlifeThe biggest gap to close here. PR into these lists on a specific, honestly-owned angle.
Booking.comOTA“Most booked” (8.4) but outside the ranked top-10 5-star moduleMore recent 5-star reviews to enter the ranked module.
Expedia, Trip.com, TrivagoOTA / meta-searchStandard listingsTransactional listings, with little story to shape.
The hotel’s own siteOwnedCited only for branded or comparison questionsKeep it accurate; it isn’t what wins the open questions.

Why sampling matters

The two engines disagreed on the same business, the same questions. ChatGPT’s search pass named the dental clinic in three of six runs, leaning on Google reviews, where the clinic’s rating is near-perfect. The pure web-grounded pass found that same clinic absent from every generic top-10, where results are dominated by competitors’ own SEO pages and directories the clinic isn’t on. Same business, same questions, but the answers were opposite, depending on how the engine retrieves. That’s exactly why a real scan samples multiple engines, multiple times, instead of trusting one run.

05

The reusable playbook

For any "best [category] in [city]" AI query, this is the order to close the gaps in. It generalized across both scans, and it should generalize to most local categories:

  1. Own Google first.

    A complete Business Profile, the right categories, real photos, and a steady stream of recent reviews. Every engine in this scan leaned on Google hardest.

  2. Get onto the category’s third-party directories and review platforms.

    These are the directories built specifically for the vertical: health directories for a clinic, TripAdvisor and Booking for a hotel. Claim, complete, and earn ratings there. Most businesses are only half-present.

  3. Win editorial “best of” listicles.

    This is outreach and PR, the single biggest gap in both scans, and what AI most reliably summarizes for a “best X in Doha” question. Pick an angle the business can honestly own.

  4. Build genuine community-forum presence.

    Qatar Living, Reddit, TripAdvisor forums: this is where “which one do people recommend” answers actually ground.

  5. Make the business’s own site answer-ready.

    Clear entity content, schema, and FAQs. Necessary, but not sufficient: in both scans, the owned site was cited last.

Measure it honestly: Because the two engines disagreed on the same question, the honest way to measure this is never a single run: sample each engine several times, report a mention rate rather than a yes/no, and show the receipts. That’s exactly how Awal is built to measure it.

06

What this study doesn't claim

Nothing below is invented. But a study this size has real limits. Here they are, plainly:

  • Two engines, not five. This scan ran a real ChatGPT search pass plus a web-grounded proxy. The full Awal product adds more engines and more samples per question.
  • US-biased search. The web-grounded tool indexes US-market results. A Qatar-localized query would likely surface more Qatar Living, Arabic-language, and local-media sources than this scan found.
  • Reddit was under-surfaced. Zero Reddit results came back from the web-grounded tool on these questions. That’s a gap in that tool, not proof Reddit doesn’t matter for these categories.
  • Two samples per question. This scan was cost-bounded for a research pass; the full product samples more per question for a tighter confidence interval.
  • Where a source blocked automated access (TripAdvisor and Michelin both returned 403 errors on direct fetch), the figures above came from search-result snippets instead.

07

Want to see this for your own business?

Run a free scan and see exactly which sources AI cites, and skips, when someone asks about you.

Run a free scan