Self-Serve AEO Tools vs Full-Service Agency vs In-House
Tool, in-house hire or agency for AEO and GEO? A decision table, tool categories, 12 vendor questions and a written-deliverables checklist. We sell services.
Key takeaways
- Tools measure and agencies implement. Choose by who will act on the data: a tool if you have a person who can, an agency if you do not, and a hybrid for most teams.
- AI answers change between runs. The best public study we found reported under a 1 in 100 chance of the same brand list twice, so ask any vendor how many runs sit behind each number.
- Compare vendors on engines and surface covered, how prompts are chosen, raw answer capture, run counts, export, geography and pricing basis. Prices and engine lists change often, so verify them on the vendor's page.
- Do not hire an agency to fix a site that is not indexed, to get a guarantee, or when you only need measurement. Insist on written deliverables and the right to export your data.
Want to see where you stand before you choose?
We start with a free, analyst-run look at what AI assistants say about you today, then scope any on-site and off-site work in writing. No software to buy.
In this guide
Buy a self-serve tool if your gap is measurement and someone on your team can act on the results. Hire an agency if you need pages fixed and third-party coverage earned and nobody in-house can do that work, and build in-house only if there is steady work for a dedicated owner. For most teams the practical answer is a hybrid: a tool (or a manual prompt panel) to measure, and a specialist or an internal owner to implement.
A disclosure first. Reputation Resolutions sells services, not software. We have no AEO or GEO tool to sell you, which removes one bias and leaves another: we earn money when you hire an agency. This guide should help even if you never hire us. It ranks no vendors, scores none and names no winner, and where it names tools it reports only what each vendor's own public page said on November 18, 2026. In our read of the pages that rank for these searches (editorial judgment, not a study), most were written by a company selling the option it recommends, and few explained how much AI answers vary between runs. We try to cover both gaps.
AEO (answer engine optimization) and GEO (generative engine optimization) overlap heavily, and everything below applies to both. For vocabulary, see AEO vs GEO vs SEO. Once you know you want an agency, our guide on how to choose an AI SEO agency has the vetting scorecard. This guide covers the step before.
The decision in one table
Find the row that sounds like you. These are editorial judgments.
You want the measurement and the work done, with little spare time
- Usually fits
- Hybrid: a tool to measure, a specialist to implement
- Why
- The tool shows where you stand. The specialist fixes what it finds.
You do not know what AI says about you
- Usually fits
- A manual prompt panel or a free audit first
- Why
- Learn whether there is a problem before you pay.
Small team with some SEO skill and spare hours
- Usually fits
- Tool plus an internal owner
- Why
- Measurement is the gap, and your owner can fix pages.
No SEO or PR skill in-house and no spare hours
- Usually fits
- Agency, or a time-boxed specialist project
- Why
- A dashboard nobody acts on changes nothing.
Hundreds of prompts across brands or regions
- Usually fits
- Tool, with a raw-data export
- Why
- 100 prompts run three times on four engines is 1,200 answers a month by hand.
Regulated field (health, legal, finance)
- Usually fits
- Internal owner or specialist with a compliance step
- Why
- Approval matters more than volume.
A wrong fact traced to one or two sources
- Usually fits
- A targeted fix project
- Why
- Monitoring will not correct a source. Someone has to.
You need new third-party coverage or profiles
- Usually fits
- Agency or PR specialist
- Why
- A tool cannot earn coverage.
Your site may not be crawlable or indexed
- Usually fits
- Fix the basics yourself first
- Why
- Google says an indexed, snippet-eligible page is the requirement for AI features.
Five things that decide it
Budget: compare the total cost of each path, because a subscription is the small part of the bill if someone must spend hours acting on it. Team: if nobody can rewrite pages, correct profiles and talk to publishers, a tool alone will not change what answers say. Prompt volume: a handful fits a spreadsheet, several hundred does not. Regulated industry: if every claim needs review, approval is the bottleneck. Measure or change: tools sell the first, agencies and in-house teams deliver the second.
What AEO and GEO tools actually do
Most AEO tools and GEO tools are the same thing under two names: software that asks AI engines a list of questions on a schedule and records what comes back. These categories are our own grouping, not an industry standard, and one product often spans several.
Prompt tracking and monitoring
- What it does
- Runs prompts on chosen engines and records whether you appear
- What it cannot do
- Show the prompts match real buyer questions, or return one stable answer
Citation and source logging
- What it does
- Records which pages and domains an answer cites
- What it cannot do
- Explain why they were picked, or change them
Sentiment and brand perception
- What it does
- Classifies how an answer describes you
- What it cannot do
- Judge whether a claim is true or lawful. Check a sample of automated labels
Share of voice
- What it does
- Counts how often you appear next to named competitors
- What it cannot do
- Be compared across tools, because there is no shared formula
Content, schema and page checkers
- What it does
- Scores pages, suggests edits, drafts text
- What it cannot do
- Decide what is true or on brand. Google says no special schema is needed for its AI features
Crawler and AI-referral analytics
- What it does
- Shows which AI crawlers and referrers reach your site
- What it cannot do
- Show a page was cited. A fetch is not a citation
Recommendations and agents
- What it does
- Suggests or drafts actions
- What it cannot do
- Take responsibility for them
What no tool can do
- Fix a fact at the source. If an answer repeats an error from a directory, an old article or a review, the source must be corrected or outweighed. A dashboard only shows the error.
- Earn coverage. Third-party mentions exist because someone chose to write about you.
- Rewrite pages with judgment. A tool can flag a thin page. It does not know your product, legal limits or voice.
- Control the engine or choose the fixes. Google says indexing and serving are not guaranteed, and no vendor controls what an assistant says. Prioritizing takes judgment about your business.
Tools measure noisy outputs
Most vendor comparisons skip this. An AI answer is generated text and changes between runs, so a tool's numbers are estimates from samples, and the number of runs behind each one matters.
The best public evidence we found is a January 28, 2026 study by Rand Fishkin of SparkToro with Patrick O'Donnell of Gumshoe.ai. About 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI features, roughly 60 to 100 times each (2,961 runs, November and December 2025). The report says ChatGPT and Google's AI had less than a 1 in 100 chance of giving the same brand list twice, and about a 1 in 1,000 chance of giving it in the same order. Fishkin's advice was that visibility percentage across many runs looks reasonable, and that a tool reporting a 'ranking position in AI' is misleading. Limits: O'Donnell works at an AI tracking startup, which Fishkin disclosed, volunteers used their own settings, and the study is not peer reviewed.
Two papers agree. A 2024 paper (arXiv 2408.04667) found outputs still varied across 10 runs of five models set to be deterministic. A 2026 paper (arXiv 2604.07585) argues single observations of AI search answers are unreliable and visibility should be treated as a distribution. Google says AI Overviews and AI Mode may use query fan-out and different models, and Perplexity's help page says follow-up questions build on earlier ones.
For a buyer (editorial judgment): one screenshot proves nothing, a score with no run count cannot be checked, and a small move on a small panel is probably noise. Our guide to AI search visibility metrics shows how to size a panel. The study leaves open whether developer-API answers vary like answers in the product people use, so ask each vendor which surface it measures.
What tools say about themselves
The table reports what each vendor's own public page said on November 18, 2026. We read page text and tested no product. It is not a ranking, it carries no scores, and a missing vendor was simply not read. Prices change often: Profound's page says it retired its Starter and Growth plans in September 2026. Verify every price and engine list on the vendor's site.
AthenaHQ
- Engines the page names
- 11 on Starter, including ChatGPT, Perplexity, AI Overviews, Gemini, Claude
- Lowest listed price and basis
- Free tier, 300 credits. Starter $295 a month, 1 credit per AI response
Ahrefs Brand Radar
- Engines the page names
- ChatGPT, AI Overviews, AI Mode, Gemini, Perplexity, Copilot, Claude
- Lowest listed price and basis
- Custom prompts 'starts at $50/mo'. Index 'starts at $199/mo'
Evertune
- Engines the page names
- ChatGPT, Claude, Gemini, AI Overview, AI Mode, Perplexity and others
- Lowest listed price and basis
- No price on the page we read
Goodie
- Engines the page names
- 5 models on Core, up to 14 on Enterprise
- Lowest listed price and basis
- Core $295 a month (tied to annual billing)
HubSpot AEO
- Engines the page names
- ChatGPT, Gemini, Perplexity
- Lowest listed price and basis
- $50 a month, $45 if paid annually, 25 prompts
Otterly.AI
- Engines the page names
- 7 listed. Plans cover '4 of 7' engines
- Lowest listed price and basis
- Lite 29 euros a month, 16 prompts
Peec AI
- Engines the page names
- ChatGPT, Perplexity, AI Overview, AI Mode, Gemini and more by plan. Lower plans choose 3 models
- Lowest listed price and basis
- Starter $95 a month, 50 prompts
Profound
- Engines the page names
- Up to 10 listed, including ChatGPT, Perplexity, Gemini, Claude
- Lowest listed price and basis
- Enterprise only, custom
Rankshift
- Engines the page names
- 'Track all LLMs', none named
- Lowest listed price and basis
- Prices not shown in the text we could read
Scrunch
- Engines the page names
- 8 listed, including ChatGPT, Perplexity, Gemini. Claude is an add-on
- Lowest listed price and basis
- Core $250 a month, 15,000 AI responses
Semrush AI Visibility Toolkit
- Engines the page names
- ChatGPT and Google AI Mode named
- Lowest listed price and basis
- $99 a month, 25 prompts, extra user $99
SE Ranking AEO Tool
- Engines the page names
- AI Overviews and ChatGPT now, others 'coming soon'
- Lowest listed price and basis
- No price on the page. 14-day trial
Similarweb AI Search
- Engines the page names
- ChatGPT, Gemini, Perplexity, AI Mode
- Lowest listed price and basis
- $99 a month billed annually, $129 monthly
Writesonic
- Engines the page names
- ChatGPT, Gemini, AI Overviews on lower plans
- Lowest listed price and basis
- Starter $79 a month, billed annually
Three patterns matter more than any single row. The price basis differs (prompts, credits, responses, user licenses or domains), so two tools at one price can track very different amounts. A long engine list is often wider than one plan can use. And data pipelines differ: Ahrefs says its index starts from its own keyword database, stores every response and may have limited coverage for brands with little search volume, and Evertune says it samples each prompt 100 times. We did not test these statements, but they show why 'how is the data collected' is the right question.
A tool evaluation checklist
Use these eight checks on any tool. A vendor that cannot answer in writing has told you something.
Engines covered
- What to ask
- Which engines, on which plan, and which surface: API, search-enabled API or the product itself
- Why it matters
- API answers may differ from what users see
How prompts are chosen
- What to ask
- Who writes the list, and can you import your own
- Why it matters
- A list built from the vendor's data may not match how your buyers ask
Raw answer capture
- What to ask
- Are full answer text, date, engine and every cited source stored and visible
- Why it matters
- Without raw answers you cannot audit a score
Run frequency and variance
- What to ask
- How many runs per prompt, how often, and are results rates with a sample size
- Why it matters
- One run is an anecdote
Export and exit
- What to ask
- Can you export every answer and keep history if you cancel
- Why it matters
- Your history is your baseline
Geography and persona
- What to ask
- Which countries, languages and logged-in states are simulated
- Why it matters
- An answer in one place is not the answer everywhere
Pricing basis
- What to ask
- Per prompt, credit, response, seat or domain, and what triggers add-ons
- Why it matters
- The headline price rarely equals your bill
Method disclosure
- What to ask
- Will they show how scores are calculated and the runs behind them
- Why it matters
- A score you cannot inspect is a vendor construct
In-house AEO team vs agency: what in-house really costs
In-house gives you control, product knowledge and fast approvals. The hidden costs are skills, time and variance, not just salary (editorial judgment).
- Skills. The work mixes SEO, content, structured data, PR and analytics. One hire rarely covers all of it.
- Time. A hand-run panel of 20 prompts, three runs each, with search on and off, is 120 answers per engine per month. In-house also usually still needs a tool.
- Relationships and risk. Coverage depends on publishers a new hire does not know, and the program can leave with the specialist.
How agencies fail
Each of these can be checked before you sign. Our agency guide has the full scorecard.
- Guarantees. A promised citation, mention or ranking is a warning sign, since no one controls an engine. A written guarantee about a deliverable, such as an article published on a named outlet, is a different thing.
- Black-box scores. A proprietary score with no prompts, run counts or formula cannot be checked.
- No raw answers. If reports show charts but never the answers, you cannot see what changed or why.
- Repackaged SEO. Some firms relabel a standard SEO package. Ask what is different.
- A tool login sold as a service. Ask who reads the output and who is accountable by name.
- Weak terms. Unnamed placements, unclear rights to content and data, or a long lock-in.
Twelve questions for any vendor
These work for a tool company or an agency, and differ from the ten in our agency guide, which score a proposal.
- 1. Which engines do you cover, on which surface and on which plan?
- 2. Who chooses the prompts, and can I see and edit the full list?
- 3. How many times is each prompt run, how often, and how do you report variation?
- 4. Can I export every raw answer with date, engine, prompt and cited sources, and keep it if I leave?
- 5. Which countries, languages and logged-in or logged-out states do you simulate?
- 6. How is your score calculated, and can I see the formula and the runs behind it?
- 7. Who is accountable for results, by name, and what do they do each month?
- 8. What do you not do, and what will you not promise?
- 9. Who owns the content, accounts and data, and what happens when we end the contract?
- 10. What does the price cover, and what triggers extra charges?
- 11. How do you handle claims that need legal or compliance review?
- 12. How will you separate your work from other changes, and who can I call? Ask for a dated baseline and a reference, not an anonymous before-and-after chart.
A written-deliverables checklist
Whatever you buy, get these on paper before you pay. Treat the list as a request for proposal.
- A scope with named deliverables, quantities or definitions, owners and dates. 'Optimization' alone is not a deliverable.
- A dated baseline that includes the raw answers, not only a summary.
- The prompt list, approved by you, with branded and unbranded prompts marked.
- The engines and surfaces covered, the run protocol (runs per prompt, cadence, logged-in state, countries, languages) and a monthly report with rates, sample sizes and a raw-data export.
- What is out of scope, including a plain statement that citations, rankings and traffic are not guaranteed.
- Who writes, who checks facts and who approves, including your sign-off before anything is published.
- For third-party coverage: outlets named in writing, the window and the remedy if a piece does not run.
- Ownership of content, accounts and data, an export on exit, pricing, triggers for change, the term and how to cancel.
The hybrid model
A tool for measurement and a specialist for implementation is the pattern most teams settle on (editorial judgment). It works when each side has a clear job. The tool or a manual panel measures on a schedule. The specialist diagnoses why an answer says what it says, using the cited sources, and fixes or earns third-party sources. An internal owner corrects your own pages and approves every claim.
One rule: own the tool account and the data yourself, so a specialist reads it and you can change specialists without losing history. If an agency uses its own software, get the export terms before you sign.
A worked budget example (illustrative, no prices)
Take a 60-person company whose marketing manager has about five spare hours a week, no PR contacts and a panel of 40 prompts on four engines. The numbers are invented to show the logic. This is not a case study.
First, count the work, not the subscription. The tool path costs the annual subscription plus the manager's hours acting on results, times their loaded hourly cost. If the findings need skills nobody has, you pay the subscription and change nothing.
Second, read what the tool finds. Suppose it shows five pages with outdated facts, two wrong third-party profiles and no category-outlet coverage. The manager can fix the pages and perhaps the profiles. Coverage needs publisher relationships.
Third, match each gap to a model: a tool, the manager for page and profile fixes, and a specialist for coverage as a defined project with named deliverables. Compare that total with a full-service retainer and with hiring a specialist. The cheapest sticker price is rarely the cheapest path.
When you should not hire an agency
We sell agency services, so we say this plainly.
- Your site may not be crawlable or indexed. Fix that first. Google says there are no additional requirements for AI Overviews or AI Mode beyond an indexed, snippet-eligible page, and OpenAI says a site opted out of OAI-SearchBot will not appear in ChatGPT search answers.
- You need measurement only and have staff to act. Buy a tool, or run a spreadsheet panel with our monitoring guide.
- You want a guarantee. If a firm guarantees citations, walk away.
- You have little to be cited for yet. An agency cannot invent coverage or substance. Build the real asset first.
- One wrong source is the whole problem. A targeted correction may be the right purchase.
- Your budget is below the scope. A smaller scope or our GEO checklist done in-house beats an under-resourced retainer.
What drives agency cost
We publish no price list here, because scope sets the price. The drivers are prompts and engines tracked, pages in scope, how much third-party work is included, legal or compliance review, brands, regions and languages covered, reporting depth, who is assigned and the term.
Published benchmarks are thin. The one we found with a stated sample is a First Page Sage report dated October 7, 2026. It describes 1,146 quotes from 214 agencies collected July 6 to October 2, 2026, and reports a median GEO and AEO retainer of $6,850 a month. It also says only 23% of agencies publish any GEO or AEO pricing. Weigh it carefully: First Page Sage sells GEO services, the report does not say how agencies were selected, and the sample is not described as random. Treat it as one data point, not a market price. A quote with no prompt counts, deliverables or exclusions cannot be compared with another.
What Reputation Resolutions does
Reputation Resolutions has been in business since 2013, is A+ BBB accredited, is a Google Partner and has served clients in 40+ countries. We started in reputation work, so we check what a search for your name returns first. We sell analyst-led services, not software, and nothing we sell guarantees AI citations, mentions, rankings or traffic.
The simplest next step is the free AI visibility audit. An analyst runs a fixed set of prompts across ChatGPT, Gemini, Perplexity and AI Overviews and reports what each says about you and which sources it cites. If you then want help, we scope the work in writing, as described on our AI SEO agency page. Related pages: answer engine optimization, LLM SEO and AI search optimization. If your worry is what AI says about your reputation, see AI reputation management companies.
Choose the model by who will act on the data, then hold every vendor to raw answers, run counts and written deliverables.
Frequently asked questions
What is the difference between an AEO tool and an AEO agency?+
An AEO tool is software that runs prompts on AI engines on a schedule and records the answers, usually in a dashboard. An agency is a team that does the work: fixing pages, earning coverage and reporting. A tool measures and does not decide or act. Many teams use both.
Are AEO tools worth it?+
They are worth it when you need repeatable measurement and have someone who will act on the results. They are less useful on their own for a team with no capacity to make changes. Check engine coverage, run counts, raw answer capture and export first, and remember that answers vary between runs, so the numbers are estimates.
Do I need an agency if I already have an AEO tool?+
Not necessarily. If your team can rewrite pages, correct profiles and pursue coverage, the tool may be enough. If it shows gaps nobody has the time or skill to close, a specialist can handle implementation. Keep ownership of the tool account and the data either way.
Should I build an in-house AEO team or hire an agency?+
Build in-house when there is steady work for a dedicated owner and you want long-term control and product knowledge. Hire an agency when you need speed, a wider mix of skills or specific third-party work. Many companies combine an internal owner, a tool and an agency for peaks.
Can an AEO or GEO tool fix what AI says about my brand?+
No. A tool shows what answers say and which sources they cite. Fixing an answer usually means correcting or outweighing the underlying source, which is human work. No tool or vendor can guarantee what an assistant says, and Google says indexing and serving are not guaranteed.
How accurate are AEO and GEO tracking tools?+
They estimate, because AI answers vary from run to run. SparkToro's January 2026 study reported less than a 1 in 100 chance of the same brand list twice. Ask how many runs sit behind each number, which surface the tool measures and whether you can see raw answers. Treat a single run or an 'AI rank' with caution.
How much do AEO tools cost?+
On the vendor pages we read on November 18, 2026, listed starting points for paid plans ranged from 29 euros to $295 a month, with some free tiers and enterprise tiers quoted on request. The pricing basis varies, including prompts, credits, AI responses and user licenses, and prices change often. Verify on each vendor's site.
How much does an AEO agency cost?+
It depends on scope: prompts and engines tracked, pages in scope, third-party work, compliance review and term. Published benchmarks are limited. One report from an agency, based on 1,146 quotes from 214 agencies, gave a median of $6,850 a month, but it is not described as a random sample. Compare written deliverables, not totals.
Is an AEO agency different from an SEO agency?+
They overlap. Google says SEO best practices remain relevant for AI features. An AEO or GEO firm should add things a plain SEO package does not: a prompt panel with raw answers, work on third-party sources and accuracy checks. If a proposal looks like a relabeled SEO package, ask what is different.
Can I measure AEO for free?+
Yes, at small scale. Run a fixed list of prompts on a schedule in a clean session, three times each, and log every answer in a spreadsheet. Search Console shows Google traffic, and Google counts AI feature traffic in the Web search type. A manual panel gets slow at scale, which is when a tool starts to earn its cost.
Sources & references
- Google Search Central: AI features and your website
- OpenAI: Overview of OpenAI crawlers (OAI-SearchBot and GPTBot)
- Perplexity Help Center: How does Perplexity work?
- SparkToro: AIs are highly inconsistent when recommending brands or products (January 28, 2026)
- Non-Determinism of Deterministic LLM Settings (arXiv 2408.04667)
- Don't Measure Once: Measuring Visibility in AI Search (GEO) (arXiv 2604.07585)
- First Page Sage: GEO and AEO pricing statistics (October 7, 2026), published by an agency that sells GEO
- Vendor page, checked Nov 18, 2026: HubSpot AEO
- Vendor page, checked Nov 18, 2026: Semrush AI Visibility Toolkit
- Vendor page, checked Nov 18, 2026: Ahrefs Brand Radar
- Vendor page, checked Nov 18, 2026: Profound
- Vendor page, checked Nov 18, 2026: Otterly.AI
- Vendor page, checked Nov 18, 2026: Peec AI
- Vendor page, checked Nov 18, 2026: AthenaHQ
- Vendor page, checked Nov 18, 2026: Scrunch
- Vendor page, checked Nov 18, 2026: Goodie
- Vendor page, checked Nov 18, 2026: Writesonic
- Vendor page, checked Nov 18, 2026: SE Ranking AEO Tool
- Vendor page, checked Nov 18, 2026: Similarweb AI Search Intelligence
- Vendor page, checked Nov 18, 2026: Evertune
- Vendor page, checked Nov 18, 2026: Rankshift
Prefer we handle it?
Want to see where you stand before you choose?
We start with a free, analyst-run look at what AI assistants say about you today, then scope any on-site and off-site work in writing. No software to buy.
See our AI SEO approachReady to See What AI Finds About Your Reputation?
Schedule your free, confidential consultation and we'll show you exactly where you stand across search engines, review sites, and AI platforms.
