Best AI SEO and GEO Agencies: How to Choose One (Scorecard, Questions, Red Flags)

The best AI SEO and GEO agency for you is one that can show how it works, measures results against your revenue, is honest about what nobody can guarantee and fits your size, market and budget. We do not publish a ranked list of named competitors, because we cannot verify their results or prices and would be marking our own homework. Instead, here is the scorecard, the questions and the red flags we would use to choose any agency, including us.

Disclosure: KwiqRank is a digital marketing agency in Manchester, UK and Mumbai, India, and provides the services discussed here. This article is written by an agency about choosing agencies, so apply the same scepticism to us that we recommend you apply to everyone. Earlier versions of this page ranked agencies, including KwiqRank in first place. We replaced it, because a ranking by an interested party with unverifiable claims about others is not a trustworthy basis for a purchase. Prices and claims about competitors change, and we cannot verify them.

The market for AI SEO and GEO services is young and noisy. Some agencies offer real expertise. Others have renamed old packages, or sell certainty that nobody can supply. Google published guidance on evaluating third-party SEO advice in 2026, and its guide to generative AI features lists tactics to ignore for Google Search, including llms.txt files, content "chunking" and inauthentic mentions, and warns against tools that claim internal Google metrics. That is a useful filter for any pitch. This guide gives you a way to compare agencies on evidence. For the services themselves, see our guides to SEO, GEO and AEO, AI SEO tools and choosing an SEO company in India.

What an AI SEO or GEO agency should actually do

Strip away the labels, and credible work in this area covers these things.

Area	What good looks like
SEO foundations	Technical health, indexing, site structure, content that answers real questions. Google says SEO best practices remain relevant for its AI features
Access and eligibility	Checks Googlebot access and snippet settings, the Search Console AI setting, and your policy for OAI-SearchBot, PerplexityBot and other crawlers
Content with substance	Original data, first-hand experience, named experts, fact-checked sources. Not mass-produced pages
Entity and reputation work	Consistent business information, accurate profiles, genuine reviews and earned mentions
Measurement	Search Console including the Generative AI performance report, analytics, AI referral visits, a monthly question panel and reporting on enquiries or revenue
Honest communication	Explains what is confirmed, reported or unknown, and does not promise citations

If a proposal is mostly about files, markup and tricks that Google says it ignores, or promises a particular number of citations, walk away.

The scorecard: ten criteria

Score each agency from 0 to 3 on each line, using the same questions for all. A total below 18 out of 30 suggests you should keep looking.

#	Criterion	What a 3 looks like
1	Method transparency	Explains exactly what they will change, in what order and why, and you could verify it
2	Fit to Google's guidance	Advice agrees with Google's documentation and avoids the tactics Google says to ignore
3	Evidence of results	Case studies with named clients or verifiable contacts, method stated, outcomes tied to enquiries or revenue, limits admitted
4	Relevant experience	Has worked in your sector, market and site type, and can show how
5	AI visibility approach	Has a documented prompt-panel method, understands bot access and does not promise guaranteed citations
6	Measurement and reporting	Reports on business outcomes with sources and limits. Gives you direct access to data
7	Team and accountability	You know who does the work, their experience and who answers questions
8	Commercial terms	Clear scope, price, term, exit and ownership of accounts, content and data
9	Ethics and compliance	No fake reviews or paid link schemes. Follows disclosure and privacy rules
10	Cultural and practical fit	Communication style, time zone coverage and the working relationship suit you

Questions to ask every agency

About method

1. What exactly will you change on my site in the first 30 days?
2. How will you check that Google, ChatGPT and Perplexity can reach my pages?
3. How do you decide which page owns which question, and prevent pages competing?
4. What do you do about pages that already rank? How do you avoid breaking them?
5. Where will the facts in content come from, and who checks them?

About AI visibility

6. How do you measure AI citations, given that answers vary between runs?
7. Will you show me the raw answers and sources, not only a score?
8. What do you say about llms.txt and special markup? (A good answer reflects Google's guidance that Google Search ignores such files.)

About evidence

9. Can I speak to a client with a similar situation to mine?
10. What is a recent example where the approach did not work, and what did you learn?
11. How do you separate your work's effect from seasonality, other marketing and algorithm changes?

About commercial terms

12. Who owns the content, accounts and data if we stop working together?
13. What is the minimum term, and how do I exit?
14. What is included, what costs extra and what do you not do?
15. Who will actually do the work, and what is their experience?

Red flags

Red flag	Why it matters
Guaranteed rankings, citations or traffic	No one controls Google's or any AI engine's output. Google advises caution about such promises
Claims of inside knowledge or internal Google metrics	Google says no third-party tool or provider has access to its internal ranking or AI systems
A secret technique they will not explain	You cannot verify it, and legitimate methods can be explained
Selling llms.txt, special schema or "AI files" as the fix	Google says these are not needed for Google Search
Mass-produced pages	Google says producing many pages without added value may breach its scaled content abuse policy
Buying links or mentions in bulk	Google lists buying links for ranking as link spam, and says inauthentic mentions are not helpful
Reports showing activity, not outcomes	Pages published and keywords tracked do not show business results
Control of your accounts, domain or content	You should own everything and have admin access
Long lock-in with no performance review	Locks you in even if it does not work
Pressure tactics and fake urgency	A good agency can wait while you do due diligence
Fake or unverifiable reviews and case studies	Check independently

How agencies price their work

We do not quote market prices, because they vary widely by country, scope, team and competition, and published figures date quickly. Understand the structures instead.

Model	How it works	Watch out for
Monthly retainer	Fixed fee for a defined scope of ongoing work	Unclear scope, hours or deliverables. Compare quotes on deliverables, not totals
Project fee	Fixed fee for an audit, migration or content build	Scope creep and what happens after the project
Hourly or day rate	Pay for time used	Open-ended cost. Ask for estimates and caps
Performance-based	Fee tied to results such as leads or revenue	Agreeing how results are measured. Risk of short-term tactics or arguments
Hybrid	A base fee plus a variable component	Complexity. Define metrics and caps clearly

Ask for a written proposal with scope, deliverables, reporting, term, price and what is excluded. For work in India, the UK or both, discuss currency, tax and invoicing arrangements early. See our guide to choosing an SEO company in India for market-specific points.

A fair selection process

1. Define your goal and constraints in a one-page brief: business goals, markets, budget range, timeline, current performance and tracking.
2. Build a shortlist of three to five. Use referrals, industry contacts and your own research.
3. Send the same brief and questions to each. Compare answers on the scorecard.
4. Review evidence. Ask for references and check them. Read the case studies critically.
5. Request a short paid audit or pilot where possible, so you see their work and thinking before a long commitment.
6. Meet the people who will do the work.
7. Check the contract. Ownership, term, exit, data handling and responsibilities.
8. Agree what success looks like and how you will measure it, before you start.

How KwiqRank scores itself, and where we might not fit

You should not take our word, but you can ask us the same questions. In the interest of transparency, here is what we say about ourselves, which you should verify.

• What we do. SEO, GEO and answer engine optimisation, AI-assisted PPC, content, digital PR and related web development, for clients in the UK, India and globally, from our offices in Manchester and Mumbai. See our AI SEO services, GEO services and AI PPC services.
• How we work. We start with an access and eligibility check and a baseline prompt panel, plan content with one owning page per question and report on enquiries and revenue where we have the data. We follow Google's guidance and do not promise citations or rankings.
• Where we may not fit. If you want guaranteed rankings, bulk link packages, mass-produced AI pages or a provider with decades of enterprise references behind one specific vertical, we are not the right choice. If you need a very large in-house-style team on site, or a specialist niche we have not worked in, ask us for honest references and consider others.
• What you should do. Ask us for a written proposal and references, and compare us with others on the scorecard above.

Running a small pilot before a long contract

A pilot reduces risk for both sides. A sensible design:

Element	Example
Duration	Eight to twelve weeks
Scope	Technical and eligibility audit, baseline prompt panel, upgrade of three high-value pages, reporting setup
Deliverables	A written audit, prioritised plan, upgraded pages and a measurement dashboard
Success measures	Issues fixed, indexed pages, impressions and clicks on target pages, AI panel changes and enquiries
Exit	Either side can stop at the end, and you own everything produced

Be realistic: a short pilot will not prove long-term ranking results, but it does show how an agency thinks, communicates and delivers.

How to read a case study critically

• Is the client named, or can you speak to them?
• What was the starting point, and over what period?
• Is the outcome a business outcome or only a ranking or traffic number?
• Is the method explained, including what was done and when?
• Are other factors acknowledged, such as seasonality, other campaigns, brand strength?
• Are limits and failures mentioned? A case study with no caveats is marketing.

Common mistakes when hiring

• Choosing on price alone, or on the most confident promise.
• Believing a ranked list written by an interested party.
• Not asking who will do the work.
• Skipping references.
• Signing a long contract without a review point.
• Not agreeing how success is measured.
• Allowing the agency to hold your accounts or domain.
• Expecting results in weeks.

A proposal comparison template

Ask each agency for a written proposal, then compare them in a table like this. Fill in what each says, not what you hope.

Item	Agency A	Agency B	Agency C
Scope in the first 90 days
Deliverables per month
Who does the work, with roles and experience
How AI visibility will be measured
How results will be reported, and how often
Access you will have to accounts and data
Content approach and fact-checking process
Link and mention approach, with examples
Price, billing and what is excluded
Term, exit and ownership terms
References offered
Scorecard total

How to check references properly

1. Ask for clients with a similar size, sector and goal to yours, not only the biggest names.
2. Prepare questions. What did they do? How did communication work? What results did you see in enquiries or revenue? What went wrong? Would you hire them again?
3. Ask about access. Did you own your accounts and data? How was handover?
4. Ask about change. How did the agency handle a traffic drop or an algorithm update?
5. Verify independently. Check that the client really uses the agency, and see whether the results quoted match what is visible in public data.

What an honest first month looks like

Week	What you should see
1	Account access set up. Tracking and Search Console checked. Questions about your business and customers
2	A technical and eligibility audit, including crawler access for the engines you care about. A first look at indexing and Core Web Vitals
3	A keyword and question map, with one owning page per question, and a baseline question panel across engines
4	A written plan with priorities, owners and timelines, and the first fixes shipped. A clear report on what was found, including what is not a priority

If you only see a long list of generic recommendations or a promise of results in the first month, ask what is being delivered and measured.

Agency, in-house or freelancer?

Option	Good when	Watch out for
Agency	You need a team with several skills, and capacity you do not have	Junior staff doing the work, and account manager turnover
Freelancer or consultant	You need senior expertise on a focused problem or a lighter retainer	Capacity limits and dependence on one person
In-house hire	SEO is central to your business and you can recruit and manage the person	One person cannot cover every specialism. Training and tool costs
Hybrid	You want in-house ownership with outside specialists for audits, technical work or content	Clarify who owns what, and keep reporting joined up

A worked example: scoring three proposals

This is an invented example to show how the scorecard works in practice. A Manchester ecommerce brand receives proposals from three agencies and scores each from 0 to 3 on the ten criteria above.

Criterion	Agency A	Agency B	Agency C
1. Method transparency	3: detailed 90-day plan	1: general list of services	2: clear but generic
2. Fit to Google's guidance	3: cites official documentation	1: sells llms.txt and schema packages	2: mostly sound
3. Evidence of results	2: two named references, one detailed case study	1: unnamed percentage claims	3: three references, with enquiries and revenue data
4. Relevant experience	3: several ecommerce clients	1: mainly local services	2: some ecommerce
5. AI visibility approach	3: prompt panel and bot access audit	0: promises guaranteed citations	2: panel, but no raw answers shared
6. Measurement and reporting	3: GA4, Search Console and CRM-based reports	1: rank tracker screenshots	2: monthly dashboard, limited access
7. Team and accountability	2: senior lead named, others unnamed	1: unclear	3: team bios and roles
8. Commercial terms	3: three-month pilot, clear exit	1: 12-month lock-in	2: six-month term
9. Ethics and compliance	3: no bulk links, disclosure policy	0: package of 50 links a month	3: clear policy
10. Communication and fit	2: good, some time zone gaps	2: responsive	3: strong overlap and rapport
Total	27	8	24

The brand drops Agency B immediately, and chooses between A and C. It picks A for its pilot offer and transparency, with C as a fallback, and asks A for one more reference. The numbers do not decide for the buyer, but they force a structured conversation.

What a good GEO audit deliverable contains

Section	What it should show
Access and eligibility	Googlebot, OAI-SearchBot, PerplexityBot and other crawlers: robots.txt, firewall and log findings. The Search Console AI setting status. Snippet controls
Indexing and technical health	Indexed pages versus expected, Core Web Vitals, rendering issues, duplicate and competing pages
Question map	Questions grouped by intent, each with an owning page or flagged as a gap
Baseline panel	Results across engines, with raw answers, cited sources and competitors
Content findings	Pages that lack original value, accuracy problems, outdated facts and opportunities for data or tools
Entity and reputation findings	Inconsistent business details, missing profiles, review status and how AI tools describe the business
Measurement plan	What will be tracked, from which sources, how often and with what limits
Prioritised plan	Actions ranked by impact and effort, with owners and an honest note of uncertainty

If a proposed audit does not include most of these, ask why.

Conversations that reveal quality quickly

• Ask them to disagree with something you believe. A good agency will gently correct a misunderstanding, for example about llms.txt or guaranteed citations.
• Ask what they would not do. A good answer includes tactics they avoid and clients they decline.
• Ask for a failure story. Honest providers can describe something that went wrong and what they changed.
• Ask how they handle algorithm updates. Look for calm, process-based answers instead of promises.
• Ask who is accountable if results do not materialise. Look for clear review points, not blame.

A story of an agency selection that went well because of a pilot

This is an invented composite. A mid-sized manufacturer of industrial sensors in the Midlands wanted help with SEO and with showing up when engineers asked AI tools for supplier recommendations. The marketing director created a one-page brief and asked four agencies for proposals. Using a scorecard like the one above, she narrowed the list to two, and then did something unusual: she proposed a paid eight-week pilot to both, with the same scope.

The pilot covered an access and eligibility audit, a baseline question panel across Google, ChatGPT and Perplexity, a review of the 20 most important product and application pages, one rewritten application page, a measurement plan and a short report. Both agencies agreed. The brief stated clearly that no ranking or citation outcomes would be promised, and that the output would be judged on quality of analysis, clarity of recommendations and the usefulness of the rewritten page.

Agency A delivered a thorough audit that found real problems: an incorrectly configured firewall that blocked some crawlers, duplicate application pages and a lack of technical specifications on key pages. Its baseline panel included raw answers and a clear comparison of what competitors had that the manufacturer did not. The rewritten page was precise and well organised, with specification tables and references to standards. The report was honest about uncertainty and included a list of things it would not recommend, such as creating special files.

Agency B delivered a more polished presentation, with a visibility score and a long list of general recommendations. Its findings overlapped, but with less specific evidence. The rewritten page was fluent but generic, with several claims the engineers said they could not stand behind. The report promised significant citation improvements within three months.

The director chose Agency A, signed a six-month agreement with a review at month three, retained ownership of all accounts and assets and asked for monthly reporting on enquiries from engineers as well as visibility measures. Over the following year, the manufacturer fixed its crawler access, consolidated pages, published technical guides written with its own engineers and saw improved descriptions in AI answers, along with more enquiries mentioning technical content. The director attributed much of the success to the selection process: a scorecard, direct comparison of real work and a contract that allowed either side to review progress. She noted that Agency B was not a bad firm, but its approach did not match what her buyers valued.

If you would like to talk to us

If you would like a second opinion on a proposal from another agency, or an assessment of your site against the scorecard above, contact us. We will tell you plainly what we see, including when we think you do not need to spend more.