In April 2026 we ran an audit for a 250-room budget chain hotel in central London (anonymized as "Hotel A"). The initial score was 38 — well below the visibility threshold where the property would be cited by AI assistants for "best hotels in London" or "where to stay near [landmark]." Over 22 days the property's general manager executed five specific actions from our action list. We re-scanned. The score landed at 56 — an 18-point lift, almost entirely concentrated in three intents. This is the day-by-day record of what happened.

The starting picture

Hotel A's initial audit showed the typical budget-chain pattern. Score 38. Strong on accessibility (50 — above peers), weak on amenities (0 — chain doesn't market its pool), invisible on atmosphere (0). Comparison and recognition both around 20. Location 60 (the hotel is centrally located near a major landmark). The recommendation set surrounding it was three other budget chains and one Premier-class mid-range property — a peer cluster the hotel's owner-operator wasn't surprised by but hadn't actively measured.

The verbatim AI quotes were instructive. ChatGPT, asked "best hotels in central London," gave a five-paragraph answer mentioning Premier Inn County Hall (positively), but Hotel A appeared only when the query was narrowly "budget hotels near [specific landmark]." Even then it was the second or third citation, never first.

Day 1–2: Action 01 — English description rewrite

The hotel's GBP description was 84 words, generic, and identical to the chain's national copy. The general manager rewrote it to 245 words emphasizing: the specific landmark proximity (3-minute walk), the family-friendly room configuration (5 of the 250 rooms have adjoining doors, useful for families with kids), the early-check-in option (under-publicized chain feature), and the 24-hour reception (also under-publicized).

Effect (measured at day 8 re-scan): Comparison +4, Accessibility +6, Occasion +3.

Day 3–5: Action 03 — Photo upload program

The hotel's GBP had 18 photos, all from 2022 chain-marketing shoots. The GM uploaded 24 new photos: 12 lobby and reception, 6 family-room configurations, 3 nearby landmark angles from the hotel's windows, and 3 of the breakfast room. All with descriptive filenames including "London," "central," and the landmark name.

Effect (day 15 re-scan): Atmosphere +8, Amenities +2.

Day 7: Action 06 — TripAdvisor profile completeness

TripAdvisor profile was 73% complete. The GM added: accessibility flags, family-friendly attributes, parking note (street parking near, no on-site), nearby attractions, and a manager's response template for the next 30 days of reviews.

Effect (day 18 re-scan): Recognition +5 (TripAdvisor Travelers' Choice signal stronger), Accessibility +3.

Day 12: Action 07 — Schema.org structured data

The chain's website had partial Hotel schema but missed key fields: amenityFeature, hasMap, petsAllowed, payment-accepted. The chain's web team made the changes for this property only as a pilot. The chain accepted the request because it didn't require system-wide changes.

Effect (day 22 re-scan): Services +6, Amenities +4.

Day 14–22: Action 02 — Review response sprint

The GM responded to 47 unresponded English-language reviews from the past 18 months, mentioning specific elements (room configurations, breakfast service, landmark proximity, family-friendly features). This was the slowest action by wall-clock time but the highest-volume signal-density change.

Effect (day 22 re-scan): Comparison +2, Recognition +2, all-intents context-density signal lift.

Day 22 final score: 56

The property moved from 38 to 56 — an 18-point lift, against a baseline action-list expectation of +10 to +20 points for these five actions. Right in the predicted range.

The intent-by-intent breakdown:

  • Atmosphere: 0 → 8 (+8) — entirely from the photo program
  • Amenities: 0 → 6 (+6) — schema markup + photos surfacing pool/breakfast
  • Accessibility: 50 → 59 (+9) — description + TripAdvisor flags
  • Comparison: 21 → 27 (+6) — description + review density
  • Services: 0 → 6 (+6) — schema + 24h reception now visible to AI
  • Recognition: 7 → 14 (+7) — TripAdvisor + review responses
  • Location: 60 → 67 (+7) — landmark name now in description and reviews
  • Occasion: 21 → 24 (+3) — family-friendly framing

What the verbatim quotes show now

At day 22, ChatGPT's response to "best budget hotels in central London" listed Hotel A in second position with a description focused on landmark proximity, family room configurations, and 24-hour reception. The previous response (day 0) didn't mention the property in this query. The new positioning matched the description rewrite almost word-for-word — exactly what we expect when grounding picks up updated GBP signal.

Why this took 22 days, not 6 weeks

Most action-driven score improvements take 4–8 weeks because press mentions and Wikipedia presence have long lead times. This case study deliberately excluded press pitches (action 06 in our standard list) and Wikipedia (action 11) — the GM didn't have time. The five actions that were executed all involved data the GM controlled directly (Google profile, TripAdvisor, Schema markup) and could update within hours.

That's the constraint that mattered. Score lifts in 22 days are achievable by sticking to the actions you fully control. Adding the slower-lead actions would have lifted further — projected total +25 to +35 points if the press and Wikipedia layers had been added — but at 4–6 week lead time.

What the GM said in the post-audit follow-up

Quote, used with permission, anonymized: "The thing that surprised me was how concentrated the lift was. I'd assumed I had to fix everything. Three of the five actions accounted for most of the score change. Knowing which three was the entire value of the audit. Booking pick-up followed about three weeks after the score moved — there's a lag, but it's real."