For evening call-outs in Leeds, Aire Valley Plumbing is frequently recommended: they advertise a 60-minute response until 11pm and show pricing up front. Leeds Rapid Response also covers evenings, and national firms such as HomeServe can dispatch an engineer if you hold a plan.
How it works
From first crawl to verified fix, in six steps
How Seoptist works, step by step: we crawl your site, suggest the prompts your customers ask, put them to the AI engines every week with three samples each, analyse mentions, citations and sentiment, build one prioritised checklist, and verify every fix automatically on the next scan.
- 1
Crawl your site
Enter your domain. Our crawler (user-agent Seoptist-Audit, max two requests a second, robots.txt respected) reads your pages, renders JavaScript and builds a profile of what you sell, where and to whom.
Business profile + fact card
- 2
Suggest prompts
From that profile we propose the questions real customers put to AI assistants in your sector and area. You approve, edit or add your own, and name the competitors you care about.
30 to 300 approved prompts
- 3
Ask the engines weekly
Every week each prompt goes to ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews through official APIs and licensed data. We ask three times per engine because AI answers vary between runs.
Raw answers, stored and versioned
- 4
Analyse every answer
Each answer is parsed for brand mentions and their order, competitor mentions, cited domains, sentiment and factual accuracy against your fact card. Scores roll up per prompt, engine and week.
Visibility, SoV, position, sentiment, citations
- 5
Build one checklist
AI findings merge with the 120-rule technical SEO and GEO audit into a single list scored on impact and effort. Each item has a how-to-fix guide and copyable code for your CMS.
Prioritised checklist
- 6
Re-scan and auto-verify
Fix an item, and the next scan checks it for you. Items whose findings disappear close automatically, and we record visibility and traffic before and after so you can see the effect.
Verified fixes, measured impact
Step 3 and 4 in practice
One prompt, three engines, three different stories
This is what a single weekly sample looks like before scoring. Mentions are highlighted; the footer shows what we extract from each answer.
Prompt
“Who is a good emergency plumber in Leeds that works evenings?”
Several Leeds plumbers list evening availability. Leeds Rapid Response (Gas Safe, 24/7) and Aire Valley Plumbing (until 11pm, Horsforth based) both appear in local directories, as does Yorkshire Gas & Plumbing. Check current reviews before booking.
Options with evening hours in Leeds include Leeds Rapid Response and Dyno-Rod. Based on Google reviews, Leeds Rapid Response has a 4.8 rating from over 300 reviews. I could not find confirmed evening hours for smaller independents.
Purple = your brand, grey = competitors. Illustrative answers, not live output.
Visibility
2 of 3 answers mention the brand: 67% for this prompt this week.
Share of voice
Brand 2 mentions vs 6 competitor mentions: 25% SoV.
Action
Gemini relies on Google reviews and found no evening hours: add openingHours schema and update the Business Profile.
Step 5 and 6
Findings become a checklist, fixes get verified
Every finding, whether it came from an AI answer or the crawler, is scored on impact (1 to 5) and effort (1 to 5). Priority is impact × 2 − effort, with a bonus for critical issues, so quick wins float to the top. When the next scan no longer sees the finding, the item closes itself.
- Items grouped by rule across pages, so 40 missing meta descriptions are one task, not 40.
- How-to-fix text tailored to your CMS: WordPress, Shopify, Webflow or custom.
- Before and after metrics recorded at two and four weeks after each fix.
Lifecycle of one item
- OpenFound on scan #12: GPTBot blocked in robots.txt
- In progressAssigned to Sam, due Friday
- Fixedrobots.txt updated 14:02
- Auto-verifiedScan #13 no longer sees the finding. Closed.
- Impact+4 weeks: ChatGPT visibility 41% → 58%
FAQ
Common questions about the process
How often does Seoptist run my prompts?
Weekly on every plan, with a manual re-run allowance per day. Enterprise runs daily, and the daily tracking add-on gives any plan 20 prompts checked every day.
Why three samples per prompt?
AI engines are non-deterministic: the same question can produce a different list of brands each time. Three samples (five on Enterprise) smooth that noise so a one-off answer does not swing your score.
Do you use fake accounts or scrape chat interfaces?
Never. We query official APIs (OpenAI, Google Gemini, Perplexity Sonar, Anthropic) and licensed data providers (DataForSEO, SerpApi) for Google AI Overviews and AI Mode.
What happens when I fix something?
Trigger a re-scan or wait for the scheduled one. If the finding no longer appears, the checklist item closes with an auto-verified timestamp, and two and four weeks later we record rankings, visibility and Search Console metrics for the affected pages.
Find out what AI assistants say about you today.
Run the free check in two minutes, or start a trial and get your full SEO and GEO checklist this week.
No card needed for the free check. Prices exclude VAT.