20 Questions to Ask Before Hiring an AI SEO Company

Hiring an AI SEO company needs more scrutiny than hiring a regular SEO vendor. Many sales pitches sound advanced, while the actual work remains basic SEO with new labels.

GEO, AEO, AI visibility, entity authority, prompt tracking, and answer engines can belong in a proper plan. Terms alone prove nothing. Google Search Central treats AEO and GEO as third-party labels, and frames generative AI search work as part of SEO for Google Search.

Before signing, check what will change on your site, how progress will be tracked, who will touch the account, what proof already exists, and what assets remain yours after exit.

Five checks before the sales call

• Set one business outcome for the call.
• Match agency proof to your sector.
• Compare experience with your business model.
• Check who will work on the account.
• Review scope, reporting, ownership, exit terms.

How to score the answers

Score each answer during the sales call. Use 0 for a dodge, 1 for partial detail, and 2 for a checkable answer with a task, process, report, owner, source, clause, example, or date.

Double the score for AI-search capability and proof. Weak answers in those two areas should remove a company from your shortlist.

Strategy and business match

1. What do you need to know before sending a proposal?

A proposal belongs after discovery. You need a vendor to ask how your business earns money, where enquiries come from, how sales close, and where poor leads waste staff time.

A Noida clinic, a Delhi NCR consultant, a SaaS company, and an ecommerce brand may all request AI SEO. Each business needs a different commercial plan. A complete proposal before discovery signals template work.

Use one test question: what would change in the proposal after a sales or operations call? If the answer has no substance, the plan was written too early.

2. What outcome comes first?

Pick one commercial outcome before keyword talk. You may need qualified calls, ecommerce revenue, demo requests, local enquiries, or AI answer mentions for buyer comparison prompts.

Traffic with poor intent wastes budget. AI mentions from low-value prompts create noise. A useful answer links the first outcome with page work, query groups, prompt groups, competitor checks, and conversion tracking.

High-volume keyword talk at the start is a warning. A better answer starts with the result you need from the next phase of work.

3. How do you choose the first fix?

Order reveals thinking. New content rarely belongs first. Broken tracking, thin proof, repeated service pages, weak commercial pages, or crawl problems can make fresh content wasteful.

A capable vendor starts at the bottleneck. Measurement problems get measurement work. Commercial page weakness gets page work. Crawl barriers get technical work. Missing proof gets proof work.

The answer should name one priority and the reason behind it. A service menu with no priority is packaging, not strategy.

4. Have you solved our kind of problem before?

Relevant proof beats famous proof. A large SaaS result may say little about a Delhi NCR service business. A traffic case may say little about lead quality. A normal ranking case may say little about AI answer visibility.

Look for a similar problem: weak service pages, low trust, poor local visibility, thin comparison content, missing AI mentions, poor reporting, or low-quality enquiries.

A useful case has a starting point, a decision, completed work, a time period, and an outcome. One question exposes the brochure version: what went wrong during the project?

AI-search capability

5. What does your AI SEO work include?

AI SEO must turn into visible work. The answer should name site changes, reporting changes, and proof sources.

A credible scope can cover crawl checks, indexability review, page rendering checks, service entity cleanup, schema markup review, comparison content, author proof, review hygiene, and AI visibility tracking.

No vendor controls citation on command. You are paying for access, page usefulness, source quality, wording accuracy, and measurement. If the answer is only “we optimize for ChatGPT,” ask which page, field, schema item, prompt group, or proof asset changes first.

6. What AI search surfaces deserve tracking for our business?

Your buyer journey decides the surface mix. A local business may start with Google Search features and local intent. A B2B company may need ChatGPT search or Perplexity checks for comparison prompts. An ecommerce brand may need product discovery and review signals.

A broad AI visibility score hides too much. Mention, citation, competitor appearance, answer tone, prompt coverage, and page source all carry different weight.

Ask for an anonymized report format. The client name can remain hidden. You only need to see if the report helps you choose the next content or technical move.

7. What crawler access do we want, and what tradeoff comes with it?

Crawler advice tests technical skill. OpenAI documents OAI-SearchBot and GPTBot as separate crawlers. OAI-SearchBot connects with search features, while GPTBot connects with use of crawled content for improving foundation models.

Ask what crawlers should access your site, what remains blocked, and what business tradeoff you accept. More access can support discovery. More restriction can protect control. The answer needs a reason, not trend-chasing.

A competent check covers robots.txt, blocked resources, rendering, indexability, important commercial pages, and valid page data. If the plan leans on llms.txt for Google AI visibility, ask for the Google source. Google documents no benefit or penalty from llms.txt for Google Search visibility, including generative AI features.

8. How will AI visibility be measured?

A measurement plan needs repeatability. The company should define prompt groups, surfaces, market location, competitor set, check frequency, and success signals.

Mentioned, cited, recommended, and chosen are different states. SparkToro research from January 2026 found high inconsistency in AI brand recommendations, so one screenshot fails as proof.

A practical report should show brand appearance, description accuracy, competitor presence, citation source, page support, and movement after site changes. Manual checks can work only when recorded in a format that supports comparison over time.

9. How will you build proof outside our website?

Your site is one trust surface. Reviews, partner pages, interviews, videos, directories, community answers, PR, expert mentions, and research assets can affect how your brand appears across the web.

Spam mentions create risk. Low-grade forum posts, fake reviews, and mass profile creation do no trust work.

A useful proof plan starts with assets worth earning: customer evidence, credible profiles, expert commentary, clean review practice, useful data, or reference-worthy content. Ask what someone would mention about your business without payment. Silence here exposes a weak trust plan.

Proof and track record

10. How long have you worked on AI-search visibility?

AI-search work is newer than SEO. Honest wording counts. A good vendor may have years of work in crawlability, content quality, entity accuracy, schema, and proof-building, with a shorter record in AI visibility tracking.

Be alert when a vendor talks as if AI search has a fixed playbook. Search features, answer systems, reports, and prompts all change.

You want evidence discipline. False certainty now can become poor strategy later.

11. Can you show one result with baseline, work, and time period?

A result without context has little value. Traffic can rise from seasonality, brand activity, paid campaigns, competitor decline, algorithm updates, offline marketing, or content work.

Ask for one result with the starting point, the problem found, the work completed, the time period, the outcome, and any factor outside vendor control.

Treat percentages with caution. A large percentage from a tiny base can mislead. A modest increase in qualified enquiries can have more value.

12. Can we speak with a current client?

A testimonial offers controlled proof. A reference call has more value.

A current client can tell you how work feels after the sale: reporting quality, missed deadlines, response quality, senior oversight, and willingness to change course.

Keep the reference call short. Ask what was promised, where delivery worked, and what a new client should know before signing. If confidentiality blocks a reference, request protected proof: a redacted report, anonymized case review, or sample monthly review with sensitive data removed.

13. Can you name the first fix after looking at our site?

You are testing editorial and commercial judgment, with no request for a free full audit.

A good vendor can inspect your site or Search Console screenshots and name one priority. The first fix may be technical, editorial, trust-related, local, commercial, or conversion-related.

A raw export with dozens of issues has low value without priority. Ask which fix should happen first if budget allows only one change, and why that fix affects revenue, trust, crawl access, or qualified enquiries.

People and process

14. Who owns the account?

Ask for names, roles, and decision rights. You need to know who owns strategy, technical work, content review, claim review, reporting, and calls.

Junior support can help. Senior oversight must appear where wrong choices cost money. Ask how many active accounts your lead manages. A capable lead with too many accounts becomes reactive.

Competitor conflicts also need disclosure. If the vendor serves your direct competitor in the same city and service line, you need that information before signing.

15. How does AI enter the content process?

AI can help with outlines, summaries, draft support, QA, and formatting. Risk starts when AI fills unknowns, invents proof, or publishes without human review.

Ask where the workflow allows AI and where the workflow bans AI. AI should never invent a statistic, fake a quote, create false first-hand experience, add an unverified source, or approve a page for publication.

A responsible process has one named final approver. That person checks claims, sources, usefulness, and brand voice before upload. If the sales point is “100 AI pages per month,” ask how many pass human source review.

16. What work will exist after 90 days?

Avoid ranking promises. Ask for inspectable work.

After the first month, you should see priority pages, conversion points, and reporting. After the second phase, you should see page improvements, technical fixes, internal links, schema fixes, or proof-building. Near the end of quarter one, you should have completed work, an AI visibility baseline if included, and a sharper next plan.

That is a work commitment without a ranking guarantee. If the answer is only “SEO takes time,” the company asks for patience without enough accountability.

Money, ownership, and exit

17. Will we have live access to reporting?

A monthly PDF can hide too much. Ask for access to the sources behind the report where possible: Search Console, GA4, conversion tracking, dashboards, task status, rank data, and AI visibility reports if included.

Access still needs interpretation. A senior reviewer should connect numbers with decisions. The report should show what changed, how the change affects the business, and what work follows.

If the report has screenshots, green arrows, and impressions but no leads, enquiries, sales, or qualified actions, reporting is decoration.

18. What belongs inside the retainer?

Scope needs exact wording. Ask for the monthly retainer in rupees, then ask what falls outside it.

Content can mean a brief, a draft, an edited page, upload, design, or full CMS publication. Technical SEO can mean an audit only, separate from implementation. Digital PR can mean advice only, separate from outreach.

Use one concrete test. If you need a new service page researched, written, edited, designed, uploaded, internally linked, and marked up with valid schema, what part belongs inside the retainer?

A broad answer such as “everything” needs a boundary.

19. What do we retain after exit?

Ownership must appear in the contract.

You should retain access to core business assets: website content, analytics, Search Console, GA4, dashboards created for your account, keyword research you paid for, content briefs you paid for, schema files, and final reports.

Tool subscriptions can remain with the vendor. Your data and content should remain with your business.

Ask for the ownership clause before signing. A friendly promise on a call will never protect you later.

20. Can we start with a short trial before a long contract?

A trial cannot prove every outcome. It can prove work quality, communication, reporting, and priority discipline.

For many Indian SMBs, a 90-day start or month-to-month agreement lowers risk. Trial outputs should be named before invoice one: tracking setup, audit findings, priority fixes, page improvements, AI visibility baseline, or a roadmap you can use after exit.

A long contract may be fair when setup cost is high. Ask what risk justifies the lock-in. If the answer is vague, slow the deal.

Hiring score

Score each question from 0 to 2. Double the AI-search capability group and the proof group.

Weak AI-search answers should end an AI SEO hire. Weak proof should end trust in the pitch. Weak ownership should pause the contract.

Use the score as a filter. Final choice still needs human review.

Before signing

Check five items before approval. The first 90 days should name work you can inspect. Proof should include baseline, work, and time period. AI visibility should appear per surface. The contract should show ownership. The proposal should avoid guarantees.

A 2023 GEO research paper introduced generative engine optimization and tested content changes that could improve visibility in generated answers. For hiring, the practical takeaway is direct: source quality, citations, authority signals, writing quality, and page structure carry more weight than terminology.

If the company can make the work checkable, continue the conversation. If the pitch holds more weight than the work, leave the shortlist.

Final thoughts

A good AI SEO company can show the work behind the terms.

You should know what changes, why that work deserves priority, how progress gets measured, who owns delivery, and what remains yours after exit.

Ask the 20 questions. Score the call while details are fresh. Add extra weight to AI-search capability and proof. Slow down when an answer sounds polished but you cannot check it.