Skip to content
Get Ranked On TopIllustrative case studyContent Engine plan

A fictional business. The figures are a modelled scenario, not a client result.

Tilal Physiotherapy · Physiotherapy group, 3 clinics · Amman

How an Amman physio group got AI answers to cite it in the Arabic its patients actually speak

Cited in 38% of Arabic AI answers on pain, up from 3%

12 months · Middle East & Africa · Content Engine plan · Native query research · Levantine & MSA content · Answer-first pages · Therapist authorship

01The challenge

Patients describe pain in their own words. The site answered in someone else's.

Amman patients search in Levantine Arabic, yet AI answers cited Gulf hospitals and English sources, and the site's machine-translated Arabic matched no real query.

Tilal's 28 therapists see patients in three Amman clinics, but online they had no voice. The Arabic site was 212 pages machine-translated from English in a single batch: grammatical in places, phrased like a textbook nobody in Amman reads. Ask an AI assistant why your back hurts when you sit, in the Arabic you speak at home, and the answer quoted a hospital in the Gulf or an English health portal. It never mentioned a clinic ten minutes away.

One heading, two languages that never meet

On the site · machine-translated

إدارة آلام الظهر السفلية

“Management of the lower back pains”

In the search box · as patients type it

ضهري بيوجعني لما أقعد

“My back hurts when I sit”

Not one word in common. Patients even spell “back” the Levantine way, so neither a search engine nor an AI assistant connects the two.

02The diagnosis

What the audit heard when it listened first.

Machine translation

212 pages

The whole Arabic site came from one automated batch: textbook terms where patients use everyday words, sentences built in English order, and not one page answering a question the way it is actually asked.

Finding 1 of 4

Real phrasing

61%

We gathered 1,400 query variants from Search Console, the clinics' WhatsApp inquiries and front-desk notes. 61% were Levantine dialect or Arabizi (Arabic typed in Latin letters and numbers); only 27% were Modern Standard Arabic.

Finding 2 of 4

AI answers

3%

Of 240 Arabic pain questions run through major AI assistants each month, only 3% of answers cited Tilal. Gulf hospital pages took 46% of the citations and English-language sources another 25%.

Finding 3 of 4

Trust

0 of 28

None of the 28 therapists had a profile or a byline. Pages showed no reviewer, no date and no licence details, so nothing told a reader, or a machine, that a qualified clinician stood behind the advice.

Finding 4 of 4

1,400 real query variants, by how they were phrased
  • Levantine 46%
  • Arabizi 15%
  • MSA 27%
  • English 12%

03The strategy

Four bands, all pulling the same way.

Built on our Content Engine plan: content and AI-search work for brands whose customers ask questions long before they book.

  1. 1

    Native query research

    Every topic started from the patient's words, not a keyword tool's: WhatsApp and front-desk phrasing, Search Console, and the follow-up questions AI assistants suggest. Each variant was tagged Levantine, MSA, Arabizi or English and grouped into four pain topics.

  2. 2

    Levantine & MSA content

    Therapists in Amman drafted; an Arabic health editor shaped. MSA carries the headings and clinical explanations, Levantine carries the questions, symptoms and reassurance in the voice patients use, with common Arabizi spellings noted. The 212 translated pages were rewritten, merged or redirected.

  3. 3

    Answer-first pages

    Each page opens with a short, direct answer to the question as asked, then the detail: when to rest, when to move, when to see a doctor. Question-shaped headings, FAQPage and MedicalWebPage schema, and a summary an AI assistant can quote without losing the meaning.

  4. 4

    Therapist authorship

    Every page is written or reviewed by a named Tilal therapist, with a profile, qualifications, licence details and a review date. Clinic pages link to the conditions each team treats, so the expertise is visible to people and to machines.

04Dialect stretch

Pull the band: a translated page against a native one.

Pick a pain topic and switch the page version. The band stretches with the share of real monthly searches, by volume, whose phrasing the page actually covers. The list shows each way of asking and who the AI answer cites.

0%25%50%75%100%SLACKTAUT

9%of monthly searches covered

0 of 6 AI answers cite Tilal

Lower back, Machine-translated page: 9% of searches covered; 0 of 6 AI answers cite Tilal.

Page heading · Machine-translated

إدارة آلام الظهر السفلية

“Management of the lower back pains”

Every control works with Tab, Enter and Space.

Six ways of asking · modelled monthly searches in Jordan

  • ضهري بيوجعني لما أقعد

    Not covered

    My back hurts when I sit

    Levantine1,300 / moAI cites: Gulf hospital
  • ألم أسفل الظهر عند الجلوس

    Not covered

    Lower back pain when sitting

    MSA880 / moAI cites: Gulf hospital
  • dahri byouja3ni

    Not covered

    My back hurts (in Latin letters)

    Arabizi390 / moAI cites: English source
  • شو بفيد لوجع الضهر

    Not covered

    What helps with back pain?

    Levantine720 / moAI cites: Gulf hospital
  • علاج طبيعي للظهر في عمان

    Not covered

    Physiotherapy for the back in Amman

    MSA1,600 / moAI cites: Directory
  • lower back pain physiotherapy amman

    Covered

    Typed in English

    English480 / moAI cites: English source

05Month by month

12 months, one band at a time.

Native pages published, cumulative
  1. Months 1–2

    Listen

    1,400 query variants gathered and tagged, 240 AI prompts baselined, therapists interviewed in all three clinics.

  2. Months 3–4

    First band

    Lower back first, the biggest topic: the first 14 of its 18 native pages replaced 41 translated ones, each signed by a therapist.

  3. Months 5–8

    Stretch

    Knee, then neck and shoulder, then sciatica: each cluster tested against the real query list before it went live.

  4. Months 9–10

    Hold

    Pages refreshed from new front-desk questions, links tightened between symptoms, conditions and clinics, AI answers rechecked monthly.

  5. Months 11–12

    Compound

    The last translated pages retired. Therapists keep adding answers to the questions patients bring in each week.

06Results

Taut where it matters: in the words patients use.

Arabic AI answers citing Tilal

3% → 38%

7 → 91 of 240 tracked prompts

Organic sessions from Arabic queries

+212%

1,850 → 5,770 sessions a month

Organic booked assessments / month

140 → 260

140 → 260 a month across three clinics

Native Arabic pages published

64

By topic: Lower back 18, Knee 16, Neck & shoulder 14, Sciatica 16.

Modelled scenario for a fictional business — not a client result. Rankings, AI citations and traffic are never guaranteed; see our Terms.

Share of Arabic AI answers citing TilalNative pagesTranslated site left as it was (modelled)
0%10%20%30%40%M1M3M5M7M9M11M1238%4%
Who the AI answers cited

Month 1

Month 12

  • Gulf hospital pages
  • English-language sources
  • Directories & other clinics
  • Tilal

Positions for queries as patients type them

Query, as typedMeaningPhrasingMonth 1Month 12
ضهري بيوجعني لما أقعدMy back hurts when I sitLevantine100+#3
ألم أسفل الظهر عند الجلوسLower back pain when sittingMSA#41#2
خشونة الركبة علاج طبيعيKnee osteoarthritis physiotherapyMSA#64#1
ركبتي بتطقطق لما أنزل الدرجMy knee clicks when I go down the stairsLevantine100+#4
شد عضلي بالرقبة شو الحلMuscle spasm in my neck, what's the fix?Levantine100+#3
الكتف المتجمد علاج طبيعيFrozen shoulder physiotherapyMSA#38#2
علاج عرق النسا بدون عمليةSciatica treatment without surgeryMSA#22#1
الديسك ولا عرق النسا، كيف بعرف؟Disc or sciatica: how do I tell?Levantine100+#5

Modelled Google positions in Jordan. Queries are shown exactly as typed; the meaning column is our translation.

Get a plan like this

Do your patients or customers ask in a language your site doesn't speak yet? Start with the Content Engine plan, or tell us how they ask.

All case studies