How Ignite works
We count what AI says.
Then we fix what's making them say it, and count again. Four stages, running on a loop.
A few Dubai studios come up consistently. Amber & Ash is the most frequently mentioned, followed by Melosa Candles and Olive Grove Candles. All three hand-pour soy wax in small batches and deliver across the UAE.
Illustrative. Business names are fictional.
Measure
We run your buyers' real questions across ChatGPT, Gemini and Perplexity, twenty times each, from clean sessions. AI doesn't answer the same way twice, so one check is an anecdote. Twenty is a rate.
One question · 20 runs
6 / 20
named · 30% ±20
Understand
We read your site and build a fact profile: what you do, where, for whom. Every fact links to the page it came from. Missing facts stay missing and get flagged. We never invent one to fill a gap.
Fact profile · sourced
Fix
Each fix is a proposal with a stated hypothesis: which number it should move, and why. You see the diff before it ships. You approve it. One click undoes it, any time.
Proposed change · awaiting you
Prove
Same questions, same method, same cadence. The report shows what moved and what didn't; declines set in the same size type as the wins.
Cycle 1 → cycle 4
Before
6
After
16
The numbers
Five things we count.
Mention rate
How often you're named across repeated runs of one question.
runs named ÷ runs total
Share of voice
Your slice of every business AI names in your category.
you vs. the field
Citation rate
How often your own pages are given as a source for the answer.
the lever you control most
Attribute accuracy
When it names you, whether what it says is true and current.
a defect count, not a growth metric
Coverage
How many of your questions you appear in at all, even once.
breadth across the category
Diagnosis
Three ways to lose.
"The AI doesn't recommend us" describes three unrelated problems. Each needs a different fix.
Invisible
It can't retrieve you. Blocked crawlers, JavaScript-only pages, and nothing outside your own domain confirming you exist.
Misdescribed
It finds you and gets it wrong. Old hours, a service you dropped, the wrong emirate, repeated with total confidence.
Outranked
Everything's correct and a competitor still wins. More reviews, more listings, more pages that answer directly.
The work
What a fix looks like.
Your website
We draft · you approve · it shipsStructured data +
Organization, LocalBusiness, Service and FAQ markup stating plainly what you are, where, and for whom.
Answer-shaped pages +
Content that answers a real question in its first two sentences, then supports it. AI extracts answers, not brand narratives.
Entity clarity +
One canonical name, one consistent address and phone, explicit service areas and languages.
Coverage gaps +
The service-by-location pages you never wrote, and Arabic pages at real parity with correct hreflang.
Machine access +
Server-rendered content, and a deliberate call on which AI crawlers you allow. That's a business decision, we ask first.
Everywhere else
We find the gaps · you close themProfile consistency +
Google Business Profile, Apple Maps, and the directories that carry weight in your category and market.
Reviews +
Nothing corroborates a business to a model like many recent, independent people describing it.
Knowledge graph entries +
Where your organisation is genuinely eligible, a correct entry in the open graphs models draw on.
Earned mentions +
Coverage on the sources AI actually cites for your category. Slow, unglamorous, most durable.
Cadence
The first quarter.
Free scan
A short prompt set, run once, showing where you stand. No card.
Baseline
Full prompt set sampled properly, fact profile built, gaps listed.
First fixes
Access and accuracy first, clearest mechanism, fastest read.
Re-measure
Weekly sampling, monthly report. Movement compounds over months.
Fair challenges
Questions we get.
Isn't this just SEO with a new name?
There's real overlap and we'd rather say so. The difference is the unit: SEO measures position on a results page, we measure whether a name appears in a generated answer, how often, against whom.
How long before something moves?
Accuracy problems can correct within weeks. Beating an entrenched competitor takes months, because it needs corroboration you earn rather than publish.
Does this work in Arabic?
Yes, natively. Arabic prompts are a first-class set, not a translation, buyers phrase things differently and the answers differ accordingly.
Can you guarantee ChatGPT recommends me?
No. What we can do is show you where you stand, tell you why, propose fixes with a stated hypothesis, and measure honestly whether they worked.